Emerging Trends in The Social and Behavioral Sciences

Culture, Diffusion, and Networks in Social Animals

Item

Title

Culture, Diffusion, and Networks in Social Animals

Author

Mann, Janet

Singh, Lisa

Research Area

Social Interactions

Topic

Primate Studies

Abstract

Long‐term studies of social animals provide detailed data on individual attributes, behaviors, and associations that enable us to explore cultural diffusion in networks. In this essay, we describe how network science can be used to improve our understanding of cultural and information transmission. After presenting an operational definition of culture, we discuss methodologies and research questions applicable to unweighted, weighted, and dynamic networks. We then highlight relevant studies and methods for both descriptive and predictive analyses that have been used to identify culture and social learning in animal networks. Applying and extending the techniques presented will improve our understanding of information transmission, social learning, and embedded subcultures in the context of human networks.

Related Essays

Mediation in International Conflicts (Political Science), Kyle Beardsley and Nathan Danneman

Kin‐Directed Behavior in Primates (Anthropology), Carol M. Berman

The Sexual Division of Labor (Anthropology), Rebecca Bliege Bird and Brian F. Codding

Culture and Cognition (Sociology), Karen A. Cerulo

Adaptation for Culture (Anthropology), Jill M. Church

Sexual Behavior (Anthropology), Melissa Emery Thompson

Culture and Social Networks (Sociology), Jan A. Fuhse

Food Sharing (Anthropology), Michael Gurven and Adrian V. Jaeggi

Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie

Emerging Trends in Social Network Analysis of Terrorism and Counterterrorism (Sociology), David Knoke

Reconciliation and Peace‐Making: Insights from Studies on Nonhuman Animals (Anthropology), Sonja E. Koski

Cooperative Breeding and Human Evolution (Anthropology), Karen L. Kramer

Niche Construction: Implications for Human Sciences (Anthropology), Kevin N. Laland and Michael O'Brien

Cultural Conflict (Sociology), Ian Mullins

Gestural Communication in Nonhuman Species (Anthropology), Simone Pika

Economics and Culture (Economics), Gérard Roland

Production of Culture (Sociology), Vaughn Schmutz and Candace N. Miller

Vocal Communication in Primates (Anthropology), Katie E. Slocombe

Primate Allomaternal Care (Anthropology), Stacey Tecot and Andrea Baden

Identifier

etrds0068

extracted text

Culture, Diffusion, and Networks
in Social Animals
JANET MANN and LISA SINGH

Abstract
Long-term studies of social animals provide detailed data on individual attributes,
behaviors, and associations that enable us to explore cultural diffusion in networks.
In this essay, we describe how network science can be used to improve our understanding of cultural and information transmission. After presenting an operational
definition of culture, we discuss methodologies and research questions applicable
to unweighted, weighted, and dynamic networks. We then highlight relevant studies and methods for both descriptive and predictive analyses that have been used to
identify culture and social learning in animal networks. Applying and extending the
techniques presented will improve our understanding of information transmission,
social learning, and embedded subcultures in the context of human networks.

INTRODUCTION
Our survival, success, and ability to exploit resources depend on cumulative
culture, a ubiquitous feature of human societies. Virtually every facet of our
current state was shaped by cultures past; we excel in niche construction,
perhaps to a fault (Laland & O’Brien, 2011; Rendell, Fogarty, & Laland,
2011). Cultural processes also shape nonhuman animal phenotypes, albeit
to a lesser extent than in humans. Nevertheless, animal societies enable us
to study the underlying network properties and processes that are rarely
accessible in human research and investigate the relationship between
these properties and cultural transmission. For example, long-term studies
of social mammals provide multifaceted connections (e.g., interactions,
associations, kinship, location/home range, communication) and individual
properties (i.e., genotypes and phenotypes) that only a handful of human
studies, usually traditional forager societies (e.g., Hadza foragers, Apicella,
Marlowe, Fowler, & Christakis, 2012) measure. Although we cannot interview animals, privacy laws do not protect them from frequent monitoring
such that real-time behavioral data are often available. This level of detail
Emerging Trends in the Social and Behavioral Sciences. Edited by Robert Scott and Stephen Kosslyn.
© 2015 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

allows us to explore the basic properties of cultural diffusion in networks.
Here we examine how the application of network science to social animals
informs our understanding of culture and information transmission. We
highlight relevant studies and methods and then discuss future directions for
those studying both human and animal networks. These efforts complement
those of social scientists (e.g., see Pachucki & Breiger, 2010) in identifying
theoretical and methodological approaches to network science and culture.
Before continuing, a working definition of culture which is applicable or
measurable across species is needed. In a recent influential book, Laland and
Galef invited social scientists and biologists to discuss The Question of Animal
Culture (Laland & Galef, 2009). Although definitions are fiercely contested, all
contributors agreed on two underlying properties of culture. First, the transmission process involves social learning (learning from the actions or products of others) and second, the socially learned behavior must distinguish
between groups (Laland, J. R. Kendal, & R. L. Kendal, 2009). This minimalist
definition generally works in describing animal cultures, but the challenge
of demonstrating social learning in nonexperimental settings remains.
Owing to this challenge, a number of scientists have tried to eliminate ecological and genetic explanations of behavioral differences between groups
as a way to identify social learning and hence leave “culture” as the only
remaining explanation (e.g., Krützen et al., 2005; Whiten et al., 1999). This
’elimination’ method is clearly flawed, since most social phenomena have a
combination of ecological, genetic, and epigenetic components that interact
with social factors (Kappeler, Barrett, Blumstein, & Clutton-Brock, 2013;
Laland & O’Brien, 2011) and one can never prove the null (Laland & Janik,
2006; Sargeant & Mann, 2009). For example, most socially learned traits
that have been deemed cultural in animals involve foraging (e.g., pine-cone
stripping rats, termite fishing chimpanzees, sponging dolphins), but all of
these depend not only on specific ecological conditions, but also on close kin
(typically the mother) and necessarily includes association, maternal effects,
and biased learning from kin (Aisner & Terkel, 1992; Lonsdorf, Eberly, &
Pusey, 2004; Mann et al., 2008; Mann, Stanton, Patterson, Bienenstock, &
Singh, 2012). To date, few would doubt that social, ecological, demographic,
and genetic factors interact to shape animal social networks and cultural
phenomena embedded in those networks. This multitude of intrinsic and
extrinsic factors receives less focus in human studies, possibly because we
tend to attribute social choice to human networks and biological factors to
animal networks. Still, demonstrating social learning among wild animals
is difficult. As a consequence, researchers have focused on developmental
patterns of a behavior and behavior of associates (e.g., Sargeant & Mann,
2009) or used diffusion models in networks (e.g., Franz & Nunn, 2009,
Hoppitt, Boogert, & Laland, 2010) to measure social transmission.

Culture, Diffusion, and Networks in Social Animals

3

In the last decade, social network studies in the field of animal behavior
have accelerated. For example, in three of the mainstream journals, Animal
Behaviour, Behavioral Ecology, and Behavioral Ecology and Sociobiology, there
were no network studies in 2004 or 2005, one in 2006 and by 2009, 15–21
articles were published cumulatively per year (Science Citation Index search
with keyword “social network.*” This trend has continued. Along with the
increase in animal network research, a plethora of studies began focusing
on behavioral traditions and animal culture, with the specific goal of defining culture by its social transmission properties (i.e., social learning), which
naturally led to defining the underlying properties of culture using social
network analysis.
CULTURAL ANALYSIS USING SOCIAL NETWORKS
Network science is an emerging discipline that studies network representations and predictive models as a way to both explain and predict various
physical, social and biological phenomena (Easley & Kleinberg, 2010; Newman, 2010). In cultural analysis, networks are advantageous for investigating
questions at different scales from the individual (ego networks) to groups
and the network structure as a whole, where the size of the network may
range from a few to billions of individuals. Network analysis and graph theory can be used to help explain the connection between the functionality of a
group and the behavior of different members of the group (Pinter-Wollman
et al., 2014). Further, patterns of information flow both depend on network
structure and influence network structure. Unraveling this relationship is
necessary to understand the relationship between information dissemination
and social learning, that is, cultural processes. However, network structure is
not equivalent to social transmission. To understand those processes, behavioral sampling of individuals in the network is needed. This is an area where
behavioral ecologists excel.
At the basic level, networks are just a collection of points (typically referred
to as nodes, actors, or vertices) connected by lines (typically referred to as edges,
ties, links, or arcs). For simple analyses, we may consider only a simple network in which the nodes are all the same type, for example, people, animals,
organizations, proteins, or computer systems, and the edges connect two
nodes based on a relationship between the two nodes. Example relationships
include kinship, friendship, alliance partner, professional affiliation, and
email correspondence. Social network analysis allows for multiple granularities of analysis and can be beneficial for answering macro-, meso- and
micro-level questions. Examples of the macro-level questions might concern
network density, the number of individuals and paths in the network,
and the distribution of connections. Connectivity can follow a range of

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

distributions, such as random, small-world (high clustering), regular lattice
(no clustering, low heterogeneity, low randomness, and high path lengths),
or scale-free (moderate heterogeneity and randomness—many small world
networks are also scale-free). Meso-level features include distinctiveness of
clusters, community composition, centrality or isolation of communities,
and whether local neighborhoods are tightly connected. At the micro-level,
we might be interested in identifying the information brokers, hubs or highly
connected individuals or isolates. Answering such questions can inform
descriptive and predictive models on cultural processes across micro, meso,
and macro network structures. Still, node and edge attributes (i.e., cultural
behaviors) are needed to identify, quantify and model social transmission.
In a simple network model, the edges do not show the direction of the relationship, the type of relationship or the strength of the relationship. Depending on the analysis, adding one or more of these features can improve the
depth of the analysis and remove potential bias (Singh, Bienenstock, & Mann,
2010). For example, the strength of a relationship can be shown in a network
by adding weights to each edge (Wasserman & Faust, 1994). Generally, for
social networks, weights are values between zero and one. However, negative weights can be used to represent different levels of animosity between
individuals (Newman, 2004). Weighted networks inform on strong and weak
relationships and communities, including channels of high information flow,
that is, likely paths for social information transmission. Finally, adding direction to relationships enables researchers to pose questions related to relationship reciprocity and dominance (Carrington, Scott, & Wasserman, 2005). As
social systems become more complex, network analysis becomes more useful
because of its ability to accommodate features of social complexity such as
motif analysis, hierarchies, individual recognition, and the exponential “cognitive load” faced with an increasing number of social relationships (e.g.,
Dunbar, 2012).
Figure 1 shows a small example of two networks, a simple unweighted,
binary, uni-mode on the left, and a richer weighted, directed, uni-mode network on the right. Colors are used to show clusters in the networks. The
unweighted network is sparsely connected (reducing the possible number of
paths for information flow) and has two clusters with a single edge (in red)
between the clusters. Even though this network is simple, we can still see
that the composition of the two clusters is different. The blue one has a central individual that controls information flow, while the yellow one contains
a clique within it, potentially allowing for more rapid flow of information.
Because there is only one edge between the two clusters, the potential for
information flow between clusters is reduced. The weighted, directed network is also sparse. However, because of the directionality of the edges, we

Culture, Diffusion, and Networks in Social Animals

5

0.4
0.6
0.9

0.8

0.3

0.1
0.7
0.7
(a)

0.9

0.1

0.8

0.3
0.7

(b)

Figure 1 Example networks: (a) unweighted and undirected; (b) weighted and
directed.

can see that information flows from the orange nodes to the green and purple. In addition, as weights are used to capture relationship strength, we also
see that there is a mix of strong and weak relationships through the network,
and while information flow is possible, a message or a piece of information
is unlikely to travel to all the nodes. While both types of networks are informative, relationship strength and reciprocity are important factors in cultural
diffusion.
In a recent paper, Pinter-Wollman et al. (2014) provoked behavioral ecologists to think about moving beyond descriptive analyses of observed patterns, to testing specific hypotheses and predictions regarding the function
of network structures. For example, even though patterns of behavior might
correlate with associations in a network, suggestive of social learning, that
does not explain what drives the behavior or the association. Most literature,
to date has focused on descriptive analysis because of the limited number
of techniques available for predictive analysis, particularly in the context of
more complex, dynamic networks. We now highlight descriptive and predictive approaches used in the literature for identifying and modeling structures
and groups in these different types of networks.
DESCRIPTIVE SOCIAL NETWORK ANALYSIS
Presumably, social transmission predominates in local or embedded communities in a network. Literature from physics and computer science focuses
on measures of cohesion and clustering to identify communities or subsets
of individuals that are more densely connected to each other than expected
(Girvan & Newman, 2002; Newman, 2006; Palla, Barabasi, & Vicsek, 2005,
2007, Shen, Cheng, Cai, & Hu, 2009). Different measures are used to identify
communities. For example, Girvan and Newman (2002) remove edges with
high betweeness (the fraction of shortest paths that traverse an edge) to

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

identify communities. Another approach proposed by Newman (2006)
partitions a network into one with high modularity (the fraction of edges in
a group minus the expected fraction if the network was random) to identify
communities. While these two methods propose algorithms that identify
nonoverlapping communities, Palla et al. (2005) propose using a clique
percolation method that finds maximal cliques to identify overlapping
communities. If social transmission is taking place, then these communities
that are based on network topology would also exhibit similar behavior
(potential subcultures).
While these works focused on static, binary networks, recent techniques have begun considering how communities change and evolve
over time (Backstrom, Huttenlocher, Kleinberg, & Lan, 2006; Caravelli,
Wei, Subak, Singh, & Mann, 2013; Gorke, Hartmann, & Wagner, 2009;
Tantipathananandh, Berger-Wolf, & Kempe, 2007) and how users behave
in these groups (Sharara, Singh, Getoor, & Mann, 2011, 2012). Looking at
these dynamic groups in the context of social transmission, we can measure
if “self-selection” is taking place and individuals are attracted to each other
based on socially learned traits. That is, modularity has a reciprocal nature
in networks, increasing cohesion and social transmission at the same time.
In animal networks, killer whales exhibit similar “dialects” and calls within
matrilineal units (subcommunities), clearly via shared association with kin
(e.g., Yurk, Barrett-Lennard, Ford, & Matkin, 2002). Similarly, sperm whale
matrilineal units use distinct codas that also appear to be socially learned
(Rendell & Whitehead, 2003). In both cases, the communication system is
used for cohesion. It is rare in animal societies however, that there is high
behavioral heterogeneity such that subcultural units within a larger network
can be readily identified. Typically, all members of a community engage in
the socially learned behavior (e.g., termite fishing, Whiten et al., 1999).
Community structures have been analyzed to identify probable cases of
social learning and culture in animal societies (Cantor & Whitehead, 2013).
A few have used dynamic approaches to investigate cultural transmission in
animal networks, such as lobtail foraging in humpback whales in the North
Atlantic (Allen, Weinrich, Hoppitt, & Rendell, 2013). In this case, they used
an order of acquisition analysis (see Hoppitt et al., 2010) to examine diffusion
in humpback whale networks over time. The detail on dynamic interactions between naïve and knowledgeable individuals was weak, although
the pattern over decades was strongly suggestive of social transmission.
Dynamic approaches are particularly valuable for investigation of group
structure evolution and the changing dynamics of group membership. For
example, Caravelli et al. (2013) adjust static community detection algorithms
to dynamic ones to better understand the evolution of communities over
time. The authors also develop metrics related to frequency of appearance of

Culture, Diffusion, and Networks in Social Animals

7

individuals in groups over time to better understand the longevity of social
relationships. Dynamic measures such as stability and diversity in group
participation, where stable actors are those who participate in the same
group over time, while diverse actors are those who consistently participate
across a number of different groups over time (Sharara et al., 2012) can also
serve as a tool for understanding cultural change and stability. A variety of
studies (e.g., Allen et al., 2013; Blonder, Wey, Dornhaus, James, & Sih, 2012;
Boogert, Reader, Hoppitt, & Laland, 2008) have used dynamic methods for
unveiling social transmission in networks.
Binary networks have received far more attention in human networks than
animal networks, possibly because of the view that weighted networks provide similar information as binary in terms of topology (Garlaschelli & Loffredo, 2009; Mastrandrea, Squartini, Fagiolo, & Garlaschelli, 2013; although
see Rankin et al. submitted), but also because, except in social media and
phone networks, we rarely have weighted information in human networks.
Behavioral ecologists typically collect weighted data on their subjects such
as time together or rates of interaction. Such weights are considered critical
components of information transmission (e.g., Whitehead & Lusseau, 2012)
and are presumably relevant in human societies where social relationships
span a continuum based on such factors as frequency, closeness/intimacy,
strength, importance, and valence.
A common approach for computing weights in animal social networks is
the social affinity or association indices (Whitehead, 2008). These measures
account for the number of times each individual is “sighted” alone and with
every other individual to create a ratio for each pair of individuals ranging
from 0 to 1 where 1 indicates that the pair is always together. The strength
of social affinity is that it is an asymmetric weight that maintains relationship direction, capturing individual’s relative sociability and sighting rate
independently of other individuals in the network. In other literatures, traditional community detection algorithms are adjusted to consider weights
(Newman, 2004; Opsahl & Panzarasa, 2009). For example, Newman calculates the betweeness of edges as if weights do not exist, and then divides
the betweenness by the weight of the edge before partitioning the network
into communities. Opsahl and Panzarasa propose using a generalized global
clustering coefficient as a measure to identify members of the same community. The strength of such methods depends on how the weights are initially
computed.
Another direction considers identifying key individuals involved in information transmission processes. Several studies have identified key individuals in information transmission. Some of these approaches involve actual
or modeled targeted removals to determine how information flow might

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

be disrupted or other social changes take place. In the study by Flack, Girvan, de Waal, and Krakauer (2006), removal of specific pigtail macaques that
served as “peacemakers” or “police-monkeys” in captive groups disrupted
the social structure and would presumably impact social transmission. In a
different approach, Williams and Lusseau (2006) simulated the consequences
of targeted versus random removals of killer whales in a wild population
and demonstrated that targeted removals fragmented social units and would
likely disrupt social transmission. Actual removals from culling or poaching
among African elephants can have social impacts that last for decades largely
because cultural information is lost (Archie & Chiyo, 2012; Shannon et al.,
2013).
At the heart of information transmission is determining how to model the
transmission of ideas. One direction of research investigates ways to model
the flow through a network. Similar to disease transmission, we can model
nodes in a network as active, for example, informed or influenced, or inactive.
Active nodes can then use the edges of the network to spread the contagion (e.g. information). For example, Domingos and Richardson (2001) used a
global, probabilistic model that employed the joint distribution of the behavior over all the nodes to find influential ones. Kempe, Kleinberg, & Tardos
(2003) used a diffusion process that begins with an initial set of active nodes
and used different weighting schemes to determine whether or not a neighbor should be activated.
PREDICTIVE SOCIAL NETWORK ANALYSIS
Traditional statistical methods can be difficult to use with relational data
because by definition, the members of the network are not independent from
each other. In fact, it is the relationship between individuals that is typically
of interest in the analysis. This lack of independence makes it difficult to interpret with traditional statistical tests. However, there are some methods that
are particularly exciting for hypothesis testing about the function and role of
different interaction patterns and relationships.
Relational variables tend to be central for predictive analyses that involve
networks. Two approaches that have been widely used to model dependencies between relational variables are multiple regression quadratic
assignment procedure (MRQAP) (Dekker, Krackhardt, & Snijders, 2007;
Krackhardt, 1988) and exponential random graph models (ERGM) (Snijders,
2002; Snijders, Pattison, Robins, & Handcock, 2006). Instead of variables
that are vectors of values, both methods have variables that are entries
in an adjacency matrix to capture the relational aspect of the data. Both
methods are meant for cross sectional network data analysis, however, they

Culture, Diffusion, and Networks in Social Animals

9

are designed to answer slightly different questions. (The discussion below is
based on the ones presented in Dekker et al., (2007) and Snijders (2002).)
MRQAP was designed to investigate factors affecting pairwise associations.
Given two different square matrices, is there an association between the same
entries in these different matrices? MRQAP is a relational version of a standard regression analysis that can be used on weighted or binary networks.
The procedure itself is a set of permutation tests for multiple linear regression model coefficients. The determinants are at the relationship or dyad level
and the dependencies generated by the network structure as a whole are controlled for. Questions of interest are typically of the form—How do different
factors (age, gender, etc.) influence the strength of association in undirected,
directed, weighted or binary dependent variables?
We recently applied the MRQAP to investigate whether individuals group
because they share a cultural trait—in contrast to animals sharing a cultural
trait because they group (e.g., killer whale dialects, Yurk et al., 2002). We study
sponge tool use in wild bottlenose dolphins where the dolphins (spongers)
use basket shaped marine sponges to ferret prey from the seafloor (Mann
et al., 2008; Patterson & Mann, 2011). Only about 5% of community members
use sponges in this way (Mann & Patterson, 2013), but spongers associate
regularly with nonspongers (Mann et al., 2012). Our recent work demonstrated that a subset of the community preferentially associate based on their
tool-using status, when sex, location, and maternal kinship are controlled for
(Mann et al., 2012). Because dolphins learn “sponging” from their mothers
and tend to be solitary while hunting with their sponge tools (Mann et al.,
2008), we were able to show homophily based on tool-using status, going a
step beyond others have shown to date. That is, spongers appear to prefer to
be with each other, suggesting that they identify with others similar to themselves (Mann et al., 2012). This is similar to patterns found in human social
groups (McPherson, Smith-Lovin, & Cook, 2001). Homophily based on similar age, sex, reproductive state or other factors is common in animals, but
they usually share the cultural trait because they group, not group based on
shared cultural traits (Mann et al., 2012).
In contrast, ERGM is designed to model networks as a whole, considering
dependencies between different relational variables jointly. This allows one
to model structural dependencies in the network. ERGM can still be used
to model dependencies between relational variables while controlling for
network structure, but it is well suited for modeling structural dependencies
of the network. In an ERGM analysis, relationships can be directed or
undirected, but the relationships must be binary. An ERGM analysis uses
stochastic modeling to determine the probability that a connection exists
between individuals based on some set of predictor variables. The explanatory variables may be attributes of the individuals, dyadic, or network

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

features. Questions of interest are typically of the form—How do various
factors influence the structure of binary networks? Henrich and Broesch
(2011) used ERGM to study transmission processes of critical cultural
information on medicinal plants, fishing, and yam farming among Fijian
villagers and had fascinating results. They found, for example, that social
learning biases were stronger towards those who demonstrated success at
a given behavior (fishing and farming) than knowledge per se (Henrich &
Broesch, 2011). ERGM is used extensively in the social sciences, but has
recently been applied to animal networks although not in the domain of
culture (Pinter-Wollman et al., 2014; Ilany, Barocas, Koren, Kam, & Geffen,
2013). One limitation with ERGM is that weighted approaches have only
recently been developed (Krivitsky, 2012).
SUMMARY AND NEW DIRECTIONS
As more people share behavioral information online publicly, researchers
will have the opportunity to better understand human behavior and the
influence of social relationships on this behavior. This essay presented
different studies and methods that have been proposed for identifying
culture in animal societies. Applying and extending both the descriptive
and predictive technique presented will improve our understanding of
information transmission and social learning in the context of human
networks. This in turn may help researchers identify subcultures that are
embedded in human networks.
As mentioned previously, many of the methods for descriptive analysis do
not consider complex networks that contain multiple node types, multiple
edge types, weights, reciprocity, and attributes. Including all of these network features is important for more complete descriptive analysis. Similar
extensions are needed for inference models as well. Without them, our ability
to answer questions related to culture and social learning will remain limited.
Acknowledgments: We are grateful to our colleagues on The Shark Bay
Dolphin Research Project and our funding sources: NSF, 0941487, 0918308,
ONR 10230702, the National Geographic Society Committee for Research and
Exploration and Georgetown University.
REFERENCES
Aisner, R., & Terkel, J. (1992). Ontogeny of pine cone opening behaviour in the black
rat, Rattus rattus. Animal Behaviour, 44, 327–336.
Allen, J., Weinrich, M., Hoppitt, W., & Rendell, L. (2013). Network-based diffusion
analysis reveals cultural transmission of lobtail feeding in humpback whales. Science, 340(6131), 485–488.

Culture, Diffusion, and Networks in Social Animals

11

Apicella, C. L., Marlowe, F. W., Fowler, J. H., & Christakis, N. A. (2012). Social networks and cooperation in hunter-gatherers. Nature, 481(7382), 497–501.
Archie, E. A., & Chiyo, P. I. (2012). Elephant behaviour and conservation: Social relationships, the effects of poaching, and genetic tools for management. Molecular
Ecology, 21(3), 765–778.
Bacher, K., Allen, S., Lindholm, A. K., Bejder, L., & Krützen, M. (2010). Genes or
culture: Are mitochondrial genes associated with tool use in bottlenose dolphins
(Tursiops sp.)? Behavior Genetics, 40(5), 706–714.
Backstrom, L., Huttenlocher, D., Kleinberg, J., & Lan, X. (2006). Group formation in
large social networks: Membership, growth, and evolution. International Conference for Knowledge Discovery and Database (KDD), pp. 44–54.
Blonder, B., Wey, T., Dornhaus, A., James, R., & Sih, A. (2012). Temporal dynamics
and network analysis. Methods in Ecology and Evolution, 3, 958–972.
Boogert, N. J., Reader, S. M., Hoppitt, W., & Laland, K. N. (2008). The origin and
spread of innovations in starlings. Animal Behaviour, 75(4), 1509–1518.
Cantor, M., & Whitehead, H. (2013). The interplay between social networks and culture: Theoretically and among whales and dolphins. Philosophical Transactions of
the Royal Society B, 368, 20120340.
Caravelli, P., Wei, Y., Subak, D., Singh, L., & Mann, J. (2013). Understanding evolving
group structures in time-varying network. Advances in Social Networks Analysis and
Mining, 142–148.
Carrington, P., Scott, J., & Wasserman, S. (2005). Models and methods in social network
analysis. New York, NY: Cambridge University Press.
Dekker, D., Krackhardt, D., & Snijders, T. A. (2007). Sensitivity of MRQAP tests to
collinearity and autocorrelation conditions. Psychometrika, 72, 563–581.
Domingos, P. & Richardson, M. (2001). Mining the network value of customers. ACM
International Conference on Knowledge Discovery and Data Mining.
Dunbar, R. I. (2012). The social brain meets neuroimaging. Trends in Cognitive Sciences,
16(2), 101–102.
Easley, D., & Kleinberg, J. (2010). Networks, crowds, and markets: Reasoning about a
highly connected world. New York, NY: Cambridge University Press.
Flack, J. C., Girvan, M., de Waal, F. B., & Krakauer, D. C. (2006). Policing stabilizes
construction of social niches in primates. Nature, 439, 426–429.
Franz, M., & Nunn, C. L. (2009). Network-based diffusion analysis: a new method
for detecting social learning. Proceedings of the Royal Society B: Biological Sciences,
276, 1829–1836.
Garlaschelli, D., & Loffredo, M. I. (2009). Generalized bose-fermi statistics and
structural correlations in weighted networks. Physical Review Letters, 102(3),
038701.
Girvan, M., & Newman, M. E. J. (2002). Community structure in social and biological
networks. Proceedings of the National Academy of Sciences, 99(2), 7821–7826.
Gorke, R., Hartmann, T., & Wagner, D. (2009). Dynamic graph clustering using
minimum-cut trees. Proceedings of the 11th International Symposium on Algorithms and Data Structures (WADS ’09), pp. 339–350.

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Henrich, J., & Broesch, J. (2011). On the nature of cultural transmission networks: Evidence from Fijian villages for adaptive learning biases. Philosophical Transactions of
the Royal Society B: Biological Sciences, 366(1567), 1139–1148.
Hoppitt, W., Boogert, N. J., & Laland, K. N. (2010). Detecting social transmission in
networks. Journal of Theoretical Biology, 263(4), 544–555.
Ilany, A., Barocas, A., Koren, L., Kam, M., & Geffen, E. (2013). Structural balance in
the social networks of a wild mammal. Animal Behaviour, 85(6), 1397–1405.
Kappeler, P. M., Barrett, L., Blumstein, D. T., & Clutton-Brock, T. H. (2013). Constraints and flexibility in mammalian social behaviour: Introduction and synthesis. Philosophical Transactions of the Royal Society B: Biological Sciences, 368(1618),
20120337.
Kempe, D., Kleinberg, J., & Tardos, E. (2003). Maximizing the spread of influence
through a social network. ACM International Conference on Knowledge Discovery and Data Mining, pp. 137–146.
Krackhardt, D. (1988). Predicting with networks—Nonparametric multipleregression analysis of dyadic data. Social Networks, 10(4), 359–381.
Krivitsky, P. N. (2012). Exponential-family random graph models for valued networks. Electronic Journal of Statistics, 6, 1100–1128.
Krützen, M., Mann, J., Heithaus, M. R., Connor, R. C., Bejder, L., & Sherwin, W. B.
(2005). Cultural transmission of tool use in bottlenose dolphins. Proceedings of the
National Academy of Sciences of the United States of America, 102(25), 8939–8943.
Laland, K. N., & Galef, B. G. (Eds.) (2009). The question of animal culture. Cambridge,
MA: Harvard University Press.
Laland, K. N., & Janik, V. M. (2006). The animal cultures debate. Trends in Ecology &
Evolution, 21, 542.
Laland, K. N., Kendal, J. R., & Kendal, R. L. (2009). Animal culture: Problems and
solutions. In K. N. Laland & B. G. Galef Jr., (Eds.), The question of animal culture
(pp. 174–197). Cambridge, MA: Harvard University Press.
Laland, K. N., & O’Brien, M. J. (2011). Cultural niche construction: An introduction.
Biological Theory, 6(3), 191–202.
Lonsdorf, E. V., Eberly, L. E., & Pusey, A. E. (2004). Sex differences in learning in
chimpanzees. Nature, 428(6984), 715–716.
Mann, J., & Patterson, E. M. (2013). Tool use by aquatic animals. Philosophical Transactions of the Royal Society., 368, 20120424.
Mann, J., Sargeant, B. L., Watson-Capps, J., Gibson, Q., Heithaus, M. R., Connor, R.
C., & Patterson, E. (2008). Why do dolphins carry sponges? PLoS One., 3(12), e3868.
Mann, J., Stanton, M. A., Patterson, E. M., Bienenstock, E. J., & Singh, L. O. (2012).
Social networks reveal cultural behaviour in tool-using dolphins. Nature Communications, 3, 980.
Mastrandrea, R., Squartini, T., Fagiolo, G., & Garlaschelli, D. (2013). Enhanced network reconstruction from irreducible local information. arXiv preprint arXiv:
1307.2104.
McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily
in social networks. Annual Review of Sociology, 27, 415–444.

Culture, Diffusion, and Networks in Social Animals

13

Newman, M. E. J. (2004). Analysis of weighted networks. Physics Review E., 70(5),
056131.
Newman, M. E. J. (2006). Modularity and community structure in networks. Proceedings of the National Academy of Sciences, 103(22), 8577–8582.
Newman, M. E. J. (2010). Networks: An introduction. New York, NY: Oxford University
Press, Inc.
Opsahl, T., & Panzarasa, P. (2009). Clustering in weighted networks. Social Networks,
31(2), 155–163.
Pachucki, M. A., & Breiger, R. L. (2010). Cultural holes: Beyond relationality in social
networks and culture. Annual Review of Sociology, 36, 205–224.
Palla, G., Barabasi, A.-L., & Vicsek, T. (2007). Quantifying social group evolution.
Nature, 446(7136), 664–667.
Palla, G., Derenyi, I., Farkas, I., & Vicsek, T. (2005). Uncovering the overlapping community structure of complex networks in nature and society. Nature, 435, 814.
Patterson, E. M., & Mann, J. (2011). The ecological conditions that favor tool use and
innovation in wild bottlenose dolphins (Tursiops sp.). PloS One, 6(7), e22243.
Pinter-Wollman, N., Hobson, E.A. Smith, J.E., Edelman, A.J. Shizuka, D., de Silva, S.,
… , McDonald, D.B. (2014). The dynamics of animal social networks: Analytical,
conceptual, and theoretical advances. Behavioral Ecology, 25(2), 242–255.
Rankin, R., Mann, J., Singh, L. O., Patterson, E. M., Krzyszczyk, E. B., & Bejder, L. (in
review). Dolphin social structure driven by weighted and topological information:
A null model approach. Animal Behaviour.
Rendell, L., Fogarty, L., & Laland, K. N. (2011). Runaway cultural niche construction. Philosophical Transactions of the Royal Society B: Biological Sciences, 366(1566),
823–835.
Rendell, L. E., & Whitehead, H. (2003). Vocal clans in sperm whales (Physeter macrocephalus). Proceedings of the Royal Society of London. Series B: Biological Sciences,
270(1512), 225–231.
Sargeant, B. L., & Mann, J. (2009). Developmental evidence for foraging traditions in
wild bottlenose dolphins. Animal Behaviour, 78(3), 715–721.
Shannon, G., Slotow, R., Durant, S. M., Sayialel, K. N., Poole, J., Moss, C., & McComb,
K. (2013). Effects of social disruption in elephants persist decades after culling.
Frontiers in Zoology, 10(1), 62.
Sharara, H., Singh, L., Getoor, L., & Mann, J. (2011). Understanding actor loyalty to
event-based groups in affiliation networks. Social Network Analysis and Mining, 1,
115–126.
Sharara, H., Singh, L., Getoor, L., & Mann, J. (2012). Finding prominent actors in
dynamic affiliation networks. Human Journal, 1(1), 1–14.
Shen, H., Cheng, X., Cai, K., & Hu, M.-B. (2009). Detect overlapping and hierarchical
community structure in networks. Physica A: Statistical Mechanics and its Applications., 388(8), 1706–1712.
Singh, L. O., Bienenstock, E. J., & Mann, J. (2010). What are we missing? Perspectives on social network analysis for observational scientific data. In B. Furht (Ed.),
Handbook of social networks: Technologies and applications (pp. 147–168). New York,
NY: Springer Science & Business Media, LLC, Chapter 7.

14

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Snijders, T. A. B. (2002). Markov chain Monte Carlo estimation of exponential random graph models. Journal of Social Structure, 3, 2.
Snijders, T. A. B., Pattison, P., Robins, G. L., & Handcock, M. (2006). New specifications for exponential random graph models. Sociological Methodology, 36(1),
99–153.
Tantipathananandh, C., Berger-Wolf, T., & Kempe, D. (2007). A framework for community identification in dynamic social networks. International Conference for
Knowledge Discovery and Database (KDD), pp. 717–726.
Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications
(Vol. 24). Cambridge, England: Cambridge University Press.
Waters, J. S., & Fewell, J. H. (2012). Information processing in social insect networks.
PLoS One, 7, e40337.
Whitehead, H. (2008). Analyzing animal societies: Quantitative methods for vertebrate
social analysis. Chicago, IL: University of Chicago Press.
Whitehead, H., & Lusseau, D. (2012). Animal social networks as substrate for cultural
behavioural diversity. Journal of Theoretical Biology, 294, 19–28.
Whiten, A., Goodall, J., McGrew, W. C., Nishida, T., Reynolds, V., Sugiyama, Y., … ,
Boesch, C. (1999). Cultures in chimpanzees. Nature. 399(6737), 682–685.
Williams, R., & Lusseau, D. (2006). A killer whale social network is vulnerable to
targeted removals. Biology Letters, 2(4), 497–500.
Yurk, H., Barrett-Lennard, L., Ford, J. K. B., & Matkin, C. O. (2002). Cultural transmission within maternal lineages: Vocal clans in resident killer whales in southern
Alaska. Animal Behaviour, 63(6), 1103–1119.

FURTHER READING
Cantor M, Whitehead H. (2013). The interplay between social networks and culture: Theoretically and among whales and dolphins. Philosophical Transactions of
the Royal Society B 368: 20120340. http://dx.doi.org/10.1098/rstb.2012.0340
Hoppitt, W., Boogert, N. J., & Laland, K. N. (2010). Detecting social transmission in
networks. Journal of Theoretical Biology, 263(4), 544–555.
Laland, K. N., & Galef, B. G. (Eds.) (2009). The question of animal culture. Cambridge,
MA: Harvard University Press.
Pinter-Wollman, N., E. A. Hobson, J. E. Smith, A. J. Edelman, D. Shizuka, S. de
Silva, … , McDonald, D. B. (2014). The dynamics of animal social networks: Analytical, conceptual, and theoretical advances Behavioral Ecology, 25(2), 242–255.
doi:10.1093/beheco/art047

JANET MANN SHORT BIOGRAPHY
Janet Mann, Professor of Biology and Psychology and Vice Provost for
Research at Georgetown University, earned her PhD at The University of
Michigan with expertise is in the field of animal behavior. Since 1988 her

Culture, Diffusion, and Networks in Social Animals

15

work has focused on social networks, female reproduction, calf development, life history, conservation, tool-use, social learning and culture among
bottlenose dolphins in Shark Bay, Australia. Her long-term study “The Shark
Bay Dolphin Research Project,” tracks over 1600 dolphins throughout their
lives. Mann has published over 80 scientific papers in journals such as Nature
Communications, Philosophical Transactions of the Royal Society, Proceedings of
the National Academy of Sciences, Proceedings of the Royal Society, Biological
Conservation, and Animal Behaviour and in books such as The Question Animal
Culture, The Biology of Traditions, Rational Animals, and Primates and Cetaceans:
Field Research and Conservation of Complex Mammalian Societies. Her edited
volume, Cetacean Societies (University of Chicago Press, 2000), received
several awards. Twice she was a fellow at The Center for Advanced Study in the
Behavioral Sciences at Stanford University. Dr. Mann’s research has received
considerable media attention worldwide, including a BBC Documentary
“The Dolphins of Shark Bay” focusing on her work in 2011. In 2013, Pamela
Turner published a children’s book “The Dolphins of Shark Bay” (Houghton
Mifflin) about Dr. Mann’s research.
http://explore.georgetown.edu/people/mannj2/

LISA SINGH SHORT BIOGRAPHY
Lisa Singh, Associate Professor in Computer Science at Georgetown
University, is an expert in large-scale data mining. She received her PhD
from Northwestern University in 1999. Her research interests include:
mining social networks, data science and analytics, privacy preserving
data mining, anomaly detection, graph databases, and sampling and bias
in social networks. Her research is supported by the National Science
Foundation and the Office of Naval Research. Dr. Singh has worked
extensively with animal data sets and social media data sets. She has
collaborated with researchers across disciplines at Georgetown University (biology, anthropology, medicine, linguistics, foreign serve, etc.), as
well as the University of Maryland, the University of California—Santa
Cruz, Hewlett Packard, the Census Bureau, and Oak Ridge National
Labs. Dr. Singh also serves on organizing and program committees
of the major data mining and database conferences, including KDD,
ICDM, SIGMOD, PVLDB, and ICDE. She is also heavily involved in
initiatives involving women in computer science and computer science
in K-12 education. More information about her work can be found at:
http://cs.georgetown.edu/∼singh.

16

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

RELATED ESSAYS
Mediation in International Conflicts (Political Science), Kyle Beardsley and
Nathan Danneman
Kin-Directed Behavior in Primates (Anthropology), Carol M. Berman
The Sexual Division of Labor (Anthropology), Rebecca Bliege Bird and Brian
F. Codding
Culture and Cognition (Sociology), Karen A. Cerulo
Adaptation for Culture (Anthropology), Jill M. Church
Sexual Behavior (Anthropology), Melissa Emery Thompson
Culture and Social Networks (Sociology), Jan A. Fuhse
Food Sharing (Anthropology), Michael Gurven and Adrian V. Jaeggi
Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie
Emerging Trends in Social Network Analysis of Terrorism and Counterterrorism (Sociology), David Knoke
Reconciliation and Peace-Making: Insights from Studies on Nonhuman
Animals (Anthropology), Sonja E. Koski
Cooperative Breeding and Human Evolution (Anthropology), Karen L.
Kramer
Implications for Human Sciences (Anthropology), Kevin N. Laland and
Michael O’Brien
Cultural Conflict (Sociology), Ian Mullins
Gestural Communication in Nonhuman Species (Anthropology), Simone Pika
Economics and Culture (Economics), Gerard Roland
Production of Culture (Sociology), Vaughn Schmutz and Candace N. Miller
Vocal Communication in Primates (Anthropology), Katie E. Slocombe
Primate Allomaternal Care (Anthropology), Stacey Tecot and Andrea Baden



Culture, Diffusion, and Networks
in Social Animals
JANET MANN and LISA SINGH

Abstract
Long-term studies of social animals provide detailed data on individual attributes,
behaviors, and associations that enable us to explore cultural diffusion in networks.
In this essay, we describe how network science can be used to improve our understanding of cultural and information transmission. After presenting an operational
definition of culture, we discuss methodologies and research questions applicable
to unweighted, weighted, and dynamic networks. We then highlight relevant studies and methods for both descriptive and predictive analyses that have been used to
identify culture and social learning in animal networks. Applying and extending the
techniques presented will improve our understanding of information transmission,
social learning, and embedded subcultures in the context of human networks.

INTRODUCTION
Our survival, success, and ability to exploit resources depend on cumulative
culture, a ubiquitous feature of human societies. Virtually every facet of our
current state was shaped by cultures past; we excel in niche construction,
perhaps to a fault (Laland & O’Brien, 2011; Rendell, Fogarty, & Laland,
2011). Cultural processes also shape nonhuman animal phenotypes, albeit
to a lesser extent than in humans. Nevertheless, animal societies enable us
to study the underlying network properties and processes that are rarely
accessible in human research and investigate the relationship between
these properties and cultural transmission. For example, long-term studies
of social mammals provide multifaceted connections (e.g., interactions,
associations, kinship, location/home range, communication) and individual
properties (i.e., genotypes and phenotypes) that only a handful of human
studies, usually traditional forager societies (e.g., Hadza foragers, Apicella,
Marlowe, Fowler, & Christakis, 2012) measure. Although we cannot interview animals, privacy laws do not protect them from frequent monitoring
such that real-time behavioral data are often available. This level of detail
Emerging Trends in the Social and Behavioral Sciences. Edited by Robert Scott and Stephen Kosslyn.
© 2015 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

allows us to explore the basic properties of cultural diffusion in networks.
Here we examine how the application of network science to social animals
informs our understanding of culture and information transmission. We
highlight relevant studies and methods and then discuss future directions for
those studying both human and animal networks. These efforts complement
those of social scientists (e.g., see Pachucki & Breiger, 2010) in identifying
theoretical and methodological approaches to network science and culture.
Before continuing, a working definition of culture which is applicable or
measurable across species is needed. In a recent influential book, Laland and
Galef invited social scientists and biologists to discuss The Question of Animal
Culture (Laland & Galef, 2009). Although definitions are fiercely contested, all
contributors agreed on two underlying properties of culture. First, the transmission process involves social learning (learning from the actions or products of others) and second, the socially learned behavior must distinguish
between groups (Laland, J. R. Kendal, & R. L. Kendal, 2009). This minimalist
definition generally works in describing animal cultures, but the challenge
of demonstrating social learning in nonexperimental settings remains.
Owing to this challenge, a number of scientists have tried to eliminate ecological and genetic explanations of behavioral differences between groups
as a way to identify social learning and hence leave “culture” as the only
remaining explanation (e.g., Krützen et al., 2005; Whiten et al., 1999). This
’elimination’ method is clearly flawed, since most social phenomena have a
combination of ecological, genetic, and epigenetic components that interact
with social factors (Kappeler, Barrett, Blumstein, & Clutton-Brock, 2013;
Laland & O’Brien, 2011) and one can never prove the null (Laland & Janik,
2006; Sargeant & Mann, 2009). For example, most socially learned traits
that have been deemed cultural in animals involve foraging (e.g., pine-cone
stripping rats, termite fishing chimpanzees, sponging dolphins), but all of
these depend not only on specific ecological conditions, but also on close kin
(typically the mother) and necessarily includes association, maternal effects,
and biased learning from kin (Aisner & Terkel, 1992; Lonsdorf, Eberly, &
Pusey, 2004; Mann et al., 2008; Mann, Stanton, Patterson, Bienenstock, &
Singh, 2012). To date, few would doubt that social, ecological, demographic,
and genetic factors interact to shape animal social networks and cultural
phenomena embedded in those networks. This multitude of intrinsic and
extrinsic factors receives less focus in human studies, possibly because we
tend to attribute social choice to human networks and biological factors to
animal networks. Still, demonstrating social learning among wild animals
is difficult. As a consequence, researchers have focused on developmental
patterns of a behavior and behavior of associates (e.g., Sargeant & Mann,
2009) or used diffusion models in networks (e.g., Franz & Nunn, 2009,
Hoppitt, Boogert, & Laland, 2010) to measure social transmission.

Culture, Diffusion, and Networks in Social Animals

3

In the last decade, social network studies in the field of animal behavior
have accelerated. For example, in three of the mainstream journals, Animal
Behaviour, Behavioral Ecology, and Behavioral Ecology and Sociobiology, there
were no network studies in 2004 or 2005, one in 2006 and by 2009, 15–21
articles were published cumulatively per year (Science Citation Index search
with keyword “social network.*” This trend has continued. Along with the
increase in animal network research, a plethora of studies began focusing
on behavioral traditions and animal culture, with the specific goal of defining culture by its social transmission properties (i.e., social learning), which
naturally led to defining the underlying properties of culture using social
network analysis.
CULTURAL ANALYSIS USING SOCIAL NETWORKS
Network science is an emerging discipline that studies network representations and predictive models as a way to both explain and predict various
physical, social and biological phenomena (Easley & Kleinberg, 2010; Newman, 2010). In cultural analysis, networks are advantageous for investigating
questions at different scales from the individual (ego networks) to groups
and the network structure as a whole, where the size of the network may
range from a few to billions of individuals. Network analysis and graph theory can be used to help explain the connection between the functionality of a
group and the behavior of different members of the group (Pinter-Wollman
et al., 2014). Further, patterns of information flow both depend on network
structure and influence network structure. Unraveling this relationship is
necessary to understand the relationship between information dissemination
and social learning, that is, cultural processes. However, network structure is
not equivalent to social transmission. To understand those processes, behavioral sampling of individuals in the network is needed. This is an area where
behavioral ecologists excel.
At the basic level, networks are just a collection of points (typically referred
to as nodes, actors, or vertices) connected by lines (typically referred to as edges,
ties, links, or arcs). For simple analyses, we may consider only a simple network in which the nodes are all the same type, for example, people, animals,
organizations, proteins, or computer systems, and the edges connect two
nodes based on a relationship between the two nodes. Example relationships
include kinship, friendship, alliance partner, professional affiliation, and
email correspondence. Social network analysis allows for multiple granularities of analysis and can be beneficial for answering macro-, meso- and
micro-level questions. Examples of the macro-level questions might concern
network density, the number of individuals and paths in the network,
and the distribution of connections. Connectivity can follow a range of

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

distributions, such as random, small-world (high clustering), regular lattice
(no clustering, low heterogeneity, low randomness, and high path lengths),
or scale-free (moderate heterogeneity and randomness—many small world
networks are also scale-free). Meso-level features include distinctiveness of
clusters, community composition, centrality or isolation of communities,
and whether local neighborhoods are tightly connected. At the micro-level,
we might be interested in identifying the information brokers, hubs or highly
connected individuals or isolates. Answering such questions can inform
descriptive and predictive models on cultural processes across micro, meso,
and macro network structures. Still, node and edge attributes (i.e., cultural
behaviors) are needed to identify, quantify and model social transmission.
In a simple network model, the edges do not show the direction of the relationship, the type of relationship or the strength of the relationship. Depending on the analysis, adding one or more of these features can improve the
depth of the analysis and remove potential bias (Singh, Bienenstock, & Mann,
2010). For example, the strength of a relationship can be shown in a network
by adding weights to each edge (Wasserman & Faust, 1994). Generally, for
social networks, weights are values between zero and one. However, negative weights can be used to represent different levels of animosity between
individuals (Newman, 2004). Weighted networks inform on strong and weak
relationships and communities, including channels of high information flow,
that is, likely paths for social information transmission. Finally, adding direction to relationships enables researchers to pose questions related to relationship reciprocity and dominance (Carrington, Scott, & Wasserman, 2005). As
social systems become more complex, network analysis becomes more useful
because of its ability to accommodate features of social complexity such as
motif analysis, hierarchies, individual recognition, and the exponential “cognitive load” faced with an increasing number of social relationships (e.g.,
Dunbar, 2012).
Figure 1 shows a small example of two networks, a simple unweighted,
binary, uni-mode on the left, and a richer weighted, directed, uni-mode network on the right. Colors are used to show clusters in the networks. The
unweighted network is sparsely connected (reducing the possible number of
paths for information flow) and has two clusters with a single edge (in red)
between the clusters. Even though this network is simple, we can still see
that the composition of the two clusters is different. The blue one has a central individual that controls information flow, while the yellow one contains
a clique within it, potentially allowing for more rapid flow of information.
Because there is only one edge between the two clusters, the potential for
information flow between clusters is reduced. The weighted, directed network is also sparse. However, because of the directionality of the edges, we

Culture, Diffusion, and Networks in Social Animals

5

0.4
0.6
0.9

0.8

0.3

0.1
0.7
0.7
(a)

0.9

0.1

0.8

0.3
0.7

(b)

Figure 1 Example networks: (a) unweighted and undirected; (b) weighted and
directed.

can see that information flows from the orange nodes to the green and purple. In addition, as weights are used to capture relationship strength, we also
see that there is a mix of strong and weak relationships through the network,
and while information flow is possible, a message or a piece of information
is unlikely to travel to all the nodes. While both types of networks are informative, relationship strength and reciprocity are important factors in cultural
diffusion.
In a recent paper, Pinter-Wollman et al. (2014) provoked behavioral ecologists to think about moving beyond descriptive analyses of observed patterns, to testing specific hypotheses and predictions regarding the function
of network structures. For example, even though patterns of behavior might
correlate with associations in a network, suggestive of social learning, that
does not explain what drives the behavior or the association. Most literature,
to date has focused on descriptive analysis because of the limited number
of techniques available for predictive analysis, particularly in the context of
more complex, dynamic networks. We now highlight descriptive and predictive approaches used in the literature for identifying and modeling structures
and groups in these different types of networks.
DESCRIPTIVE SOCIAL NETWORK ANALYSIS
Presumably, social transmission predominates in local or embedded communities in a network. Literature from physics and computer science focuses
on measures of cohesion and clustering to identify communities or subsets
of individuals that are more densely connected to each other than expected
(Girvan & Newman, 2002; Newman, 2006; Palla, Barabasi, & Vicsek, 2005,
2007, Shen, Cheng, Cai, & Hu, 2009). Different measures are used to identify
communities. For example, Girvan and Newman (2002) remove edges with
high betweeness (the fraction of shortest paths that traverse an edge) to

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

identify communities. Another approach proposed by Newman (2006)
partitions a network into one with high modularity (the fraction of edges in
a group minus the expected fraction if the network was random) to identify
communities. While these two methods propose algorithms that identify
nonoverlapping communities, Palla et al. (2005) propose using a clique
percolation method that finds maximal cliques to identify overlapping
communities. If social transmission is taking place, then these communities
that are based on network topology would also exhibit similar behavior
(potential subcultures).
While these works focused on static, binary networks, recent techniques have begun considering how communities change and evolve
over time (Backstrom, Huttenlocher, Kleinberg, & Lan, 2006; Caravelli,
Wei, Subak, Singh, & Mann, 2013; Gorke, Hartmann, & Wagner, 2009;
Tantipathananandh, Berger-Wolf, & Kempe, 2007) and how users behave
in these groups (Sharara, Singh, Getoor, & Mann, 2011, 2012). Looking at
these dynamic groups in the context of social transmission, we can measure
if “self-selection” is taking place and individuals are attracted to each other
based on socially learned traits. That is, modularity has a reciprocal nature
in networks, increasing cohesion and social transmission at the same time.
In animal networks, killer whales exhibit similar “dialects” and calls within
matrilineal units (subcommunities), clearly via shared association with kin
(e.g., Yurk, Barrett-Lennard, Ford, & Matkin, 2002). Similarly, sperm whale
matrilineal units use distinct codas that also appear to be socially learned
(Rendell & Whitehead, 2003). In both cases, the communication system is
used for cohesion. It is rare in animal societies however, that there is high
behavioral heterogeneity such that subcultural units within a larger network
can be readily identified. Typically, all members of a community engage in
the socially learned behavior (e.g., termite fishing, Whiten et al., 1999).
Community structures have been analyzed to identify probable cases of
social learning and culture in animal societies (Cantor & Whitehead, 2013).
A few have used dynamic approaches to investigate cultural transmission in
animal networks, such as lobtail foraging in humpback whales in the North
Atlantic (Allen, Weinrich, Hoppitt, & Rendell, 2013). In this case, they used
an order of acquisition analysis (see Hoppitt et al., 2010) to examine diffusion
in humpback whale networks over time. The detail on dynamic interactions between naïve and knowledgeable individuals was weak, although
the pattern over decades was strongly suggestive of social transmission.
Dynamic approaches are particularly valuable for investigation of group
structure evolution and the changing dynamics of group membership. For
example, Caravelli et al. (2013) adjust static community detection algorithms
to dynamic ones to better understand the evolution of communities over
time. The authors also develop metrics related to frequency of appearance of

Culture, Diffusion, and Networks in Social Animals

7

individuals in groups over time to better understand the longevity of social
relationships. Dynamic measures such as stability and diversity in group
participation, where stable actors are those who participate in the same
group over time, while diverse actors are those who consistently participate
across a number of different groups over time (Sharara et al., 2012) can also
serve as a tool for understanding cultural change and stability. A variety of
studies (e.g., Allen et al., 2013; Blonder, Wey, Dornhaus, James, & Sih, 2012;
Boogert, Reader, Hoppitt, & Laland, 2008) have used dynamic methods for
unveiling social transmission in networks.
Binary networks have received far more attention in human networks than
animal networks, possibly because of the view that weighted networks provide similar information as binary in terms of topology (Garlaschelli & Loffredo, 2009; Mastrandrea, Squartini, Fagiolo, & Garlaschelli, 2013; although
see Rankin et al. submitted), but also because, except in social media and
phone networks, we rarely have weighted information in human networks.
Behavioral ecologists typically collect weighted data on their subjects such
as time together or rates of interaction. Such weights are considered critical
components of information transmission (e.g., Whitehead & Lusseau, 2012)
and are presumably relevant in human societies where social relationships
span a continuum based on such factors as frequency, closeness/intimacy,
strength, importance, and valence.
A common approach for computing weights in animal social networks is
the social affinity or association indices (Whitehead, 2008). These measures
account for the number of times each individual is “sighted” alone and with
every other individual to create a ratio for each pair of individuals ranging
from 0 to 1 where 1 indicates that the pair is always together. The strength
of social affinity is that it is an asymmetric weight that maintains relationship direction, capturing individual’s relative sociability and sighting rate
independently of other individuals in the network. In other literatures, traditional community detection algorithms are adjusted to consider weights
(Newman, 2004; Opsahl & Panzarasa, 2009). For example, Newman calculates the betweeness of edges as if weights do not exist, and then divides
the betweenness by the weight of the edge before partitioning the network
into communities. Opsahl and Panzarasa propose using a generalized global
clustering coefficient as a measure to identify members of the same community. The strength of such methods depends on how the weights are initially
computed.
Another direction considers identifying key individuals involved in information transmission processes. Several studies have identified key individuals in information transmission. Some of these approaches involve actual
or modeled targeted removals to determine how information flow might

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

be disrupted or other social changes take place. In the study by Flack, Girvan, de Waal, and Krakauer (2006), removal of specific pigtail macaques that
served as “peacemakers” or “police-monkeys” in captive groups disrupted
the social structure and would presumably impact social transmission. In a
different approach, Williams and Lusseau (2006) simulated the consequences
of targeted versus random removals of killer whales in a wild population
and demonstrated that targeted removals fragmented social units and would
likely disrupt social transmission. Actual removals from culling or poaching
among African elephants can have social impacts that last for decades largely
because cultural information is lost (Archie & Chiyo, 2012; Shannon et al.,
2013).
At the heart of information transmission is determining how to model the
transmission of ideas. One direction of research investigates ways to model
the flow through a network. Similar to disease transmission, we can model
nodes in a network as active, for example, informed or influenced, or inactive.
Active nodes can then use the edges of the network to spread the contagion (e.g. information). For example, Domingos and Richardson (2001) used a
global, probabilistic model that employed the joint distribution of the behavior over all the nodes to find influential ones. Kempe, Kleinberg, & Tardos
(2003) used a diffusion process that begins with an initial set of active nodes
and used different weighting schemes to determine whether or not a neighbor should be activated.
PREDICTIVE SOCIAL NETWORK ANALYSIS
Traditional statistical methods can be difficult to use with relational data
because by definition, the members of the network are not independent from
each other. In fact, it is the relationship between individuals that is typically
of interest in the analysis. This lack of independence makes it difficult to interpret with traditional statistical tests. However, there are some methods that
are particularly exciting for hypothesis testing about the function and role of
different interaction patterns and relationships.
Relational variables tend to be central for predictive analyses that involve
networks. Two approaches that have been widely used to model dependencies between relational variables are multiple regression quadratic
assignment procedure (MRQAP) (Dekker, Krackhardt, & Snijders, 2007;
Krackhardt, 1988) and exponential random graph models (ERGM) (Snijders,
2002; Snijders, Pattison, Robins, & Handcock, 2006). Instead of variables
that are vectors of values, both methods have variables that are entries
in an adjacency matrix to capture the relational aspect of the data. Both
methods are meant for cross sectional network data analysis, however, they

Culture, Diffusion, and Networks in Social Animals

9

are designed to answer slightly different questions. (The discussion below is
based on the ones presented in Dekker et al., (2007) and Snijders (2002).)
MRQAP was designed to investigate factors affecting pairwise associations.
Given two different square matrices, is there an association between the same
entries in these different matrices? MRQAP is a relational version of a standard regression analysis that can be used on weighted or binary networks.
The procedure itself is a set of permutation tests for multiple linear regression model coefficients. The determinants are at the relationship or dyad level
and the dependencies generated by the network structure as a whole are controlled for. Questions of interest are typically of the form—How do different
factors (age, gender, etc.) influence the strength of association in undirected,
directed, weighted or binary dependent variables?
We recently applied the MRQAP to investigate whether individuals group
because they share a cultural trait—in contrast to animals sharing a cultural
trait because they group (e.g., killer whale dialects, Yurk et al., 2002). We study
sponge tool use in wild bottlenose dolphins where the dolphins (spongers)
use basket shaped marine sponges to ferret prey from the seafloor (Mann
et al., 2008; Patterson & Mann, 2011). Only about 5% of community members
use sponges in this way (Mann & Patterson, 2013), but spongers associate
regularly with nonspongers (Mann et al., 2012). Our recent work demonstrated that a subset of the community preferentially associate based on their
tool-using status, when sex, location, and maternal kinship are controlled for
(Mann et al., 2012). Because dolphins learn “sponging” from their mothers
and tend to be solitary while hunting with their sponge tools (Mann et al.,
2008), we were able to show homophily based on tool-using status, going a
step beyond others have shown to date. That is, spongers appear to prefer to
be with each other, suggesting that they identify with others similar to themselves (Mann et al., 2012). This is similar to patterns found in human social
groups (McPherson, Smith-Lovin, & Cook, 2001). Homophily based on similar age, sex, reproductive state or other factors is common in animals, but
they usually share the cultural trait because they group, not group based on
shared cultural traits (Mann et al., 2012).
In contrast, ERGM is designed to model networks as a whole, considering
dependencies between different relational variables jointly. This allows one
to model structural dependencies in the network. ERGM can still be used
to model dependencies between relational variables while controlling for
network structure, but it is well suited for modeling structural dependencies
of the network. In an ERGM analysis, relationships can be directed or
undirected, but the relationships must be binary. An ERGM analysis uses
stochastic modeling to determine the probability that a connection exists
between individuals based on some set of predictor variables. The explanatory variables may be attributes of the individuals, dyadic, or network

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

features. Questions of interest are typically of the form—How do various
factors influence the structure of binary networks? Henrich and Broesch
(2011) used ERGM to study transmission processes of critical cultural
information on medicinal plants, fishing, and yam farming among Fijian
villagers and had fascinating results. They found, for example, that social
learning biases were stronger towards those who demonstrated success at
a given behavior (fishing and farming) than knowledge per se (Henrich &
Broesch, 2011). ERGM is used extensively in the social sciences, but has
recently been applied to animal networks although not in the domain of
culture (Pinter-Wollman et al., 2014; Ilany, Barocas, Koren, Kam, & Geffen,
2013). One limitation with ERGM is that weighted approaches have only
recently been developed (Krivitsky, 2012).
SUMMARY AND NEW DIRECTIONS
As more people share behavioral information online publicly, researchers
will have the opportunity to better understand human behavior and the
influence of social relationships on this behavior. This essay presented
different studies and methods that have been proposed for identifying
culture in animal societies. Applying and extending both the descriptive
and predictive technique presented will improve our understanding of
information transmission and social learning in the context of human
networks. This in turn may help researchers identify subcultures that are
embedded in human networks.
As mentioned previously, many of the methods for descriptive analysis do
not consider complex networks that contain multiple node types, multiple
edge types, weights, reciprocity, and attributes. Including all of these network features is important for more complete descriptive analysis. Similar
extensions are needed for inference models as well. Without them, our ability
to answer questions related to culture and social learning will remain limited.
Acknowledgments: We are grateful to our colleagues on The Shark Bay
Dolphin Research Project and our funding sources: NSF, 0941487, 0918308,
ONR 10230702, the National Geographic Society Committee for Research and
Exploration and Georgetown University.
REFERENCES
Aisner, R., & Terkel, J. (1992). Ontogeny of pine cone opening behaviour in the black
rat, Rattus rattus. Animal Behaviour, 44, 327–336.
Allen, J., Weinrich, M., Hoppitt, W., & Rendell, L. (2013). Network-based diffusion
analysis reveals cultural transmission of lobtail feeding in humpback whales. Science, 340(6131), 485–488.

Culture, Diffusion, and Networks in Social Animals

11

Apicella, C. L., Marlowe, F. W., Fowler, J. H., & Christakis, N. A. (2012). Social networks and cooperation in hunter-gatherers. Nature, 481(7382), 497–501.
Archie, E. A., & Chiyo, P. I. (2012). Elephant behaviour and conservation: Social relationships, the effects of poaching, and genetic tools for management. Molecular
Ecology, 21(3), 765–778.
Bacher, K., Allen, S., Lindholm, A. K., Bejder, L., & Krützen, M. (2010). Genes or
culture: Are mitochondrial genes associated with tool use in bottlenose dolphins
(Tursiops sp.)? Behavior Genetics, 40(5), 706–714.
Backstrom, L., Huttenlocher, D., Kleinberg, J., & Lan, X. (2006). Group formation in
large social networks: Membership, growth, and evolution. International Conference for Knowledge Discovery and Database (KDD), pp. 44–54.
Blonder, B., Wey, T., Dornhaus, A., James, R., & Sih, A. (2012). Temporal dynamics
and network analysis. Methods in Ecology and Evolution, 3, 958–972.
Boogert, N. J., Reader, S. M., Hoppitt, W., & Laland, K. N. (2008). The origin and
spread of innovations in starlings. Animal Behaviour, 75(4), 1509–1518.
Cantor, M., & Whitehead, H. (2013). The interplay between social networks and culture: Theoretically and among whales and dolphins. Philosophical Transactions of
the Royal Society B, 368, 20120340.
Caravelli, P., Wei, Y., Subak, D., Singh, L., & Mann, J. (2013). Understanding evolving
group structures in time-varying network. Advances in Social Networks Analysis and
Mining, 142–148.
Carrington, P., Scott, J., & Wasserman, S. (2005). Models and methods in social network
analysis. New York, NY: Cambridge University Press.
Dekker, D., Krackhardt, D., & Snijders, T. A. (2007). Sensitivity of MRQAP tests to
collinearity and autocorrelation conditions. Psychometrika, 72, 563–581.
Domingos, P. & Richardson, M. (2001). Mining the network value of customers. ACM
International Conference on Knowledge Discovery and Data Mining.
Dunbar, R. I. (2012). The social brain meets neuroimaging. Trends in Cognitive Sciences,
16(2), 101–102.
Easley, D., & Kleinberg, J. (2010). Networks, crowds, and markets: Reasoning about a
highly connected world. New York, NY: Cambridge University Press.
Flack, J. C., Girvan, M., de Waal, F. B., & Krakauer, D. C. (2006). Policing stabilizes
construction of social niches in primates. Nature, 439, 426–429.
Franz, M., & Nunn, C. L. (2009). Network-based diffusion analysis: a new method
for detecting social learning. Proceedings of the Royal Society B: Biological Sciences,
276, 1829–1836.
Garlaschelli, D., & Loffredo, M. I. (2009). Generalized bose-fermi statistics and
structural correlations in weighted networks. Physical Review Letters, 102(3),
038701.
Girvan, M., & Newman, M. E. J. (2002). Community structure in social and biological
networks. Proceedings of the National Academy of Sciences, 99(2), 7821–7826.
Gorke, R., Hartmann, T., & Wagner, D. (2009). Dynamic graph clustering using
minimum-cut trees. Proceedings of the 11th International Symposium on Algorithms and Data Structures (WADS ’09), pp. 339–350.

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Henrich, J., & Broesch, J. (2011). On the nature of cultural transmission networks: Evidence from Fijian villages for adaptive learning biases. Philosophical Transactions of
the Royal Society B: Biological Sciences, 366(1567), 1139–1148.
Hoppitt, W., Boogert, N. J., & Laland, K. N. (2010). Detecting social transmission in
networks. Journal of Theoretical Biology, 263(4), 544–555.
Ilany, A., Barocas, A., Koren, L., Kam, M., & Geffen, E. (2013). Structural balance in
the social networks of a wild mammal. Animal Behaviour, 85(6), 1397–1405.
Kappeler, P. M., Barrett, L., Blumstein, D. T., & Clutton-Brock, T. H. (2013). Constraints and flexibility in mammalian social behaviour: Introduction and synthesis. Philosophical Transactions of the Royal Society B: Biological Sciences, 368(1618),
20120337.
Kempe, D., Kleinberg, J., & Tardos, E. (2003). Maximizing the spread of influence
through a social network. ACM International Conference on Knowledge Discovery and Data Mining, pp. 137–146.
Krackhardt, D. (1988). Predicting with networks—Nonparametric multipleregression analysis of dyadic data. Social Networks, 10(4), 359–381.
Krivitsky, P. N. (2012). Exponential-family random graph models for valued networks. Electronic Journal of Statistics, 6, 1100–1128.
Krützen, M., Mann, J., Heithaus, M. R., Connor, R. C., Bejder, L., & Sherwin, W. B.
(2005). Cultural transmission of tool use in bottlenose dolphins. Proceedings of the
National Academy of Sciences of the United States of America, 102(25), 8939–8943.
Laland, K. N., & Galef, B. G. (Eds.) (2009). The question of animal culture. Cambridge,
MA: Harvard University Press.
Laland, K. N., & Janik, V. M. (2006). The animal cultures debate. Trends in Ecology &
Evolution, 21, 542.
Laland, K. N., Kendal, J. R., & Kendal, R. L. (2009). Animal culture: Problems and
solutions. In K. N. Laland & B. G. Galef Jr., (Eds.), The question of animal culture
(pp. 174–197). Cambridge, MA: Harvard University Press.
Laland, K. N., & O’Brien, M. J. (2011). Cultural niche construction: An introduction.
Biological Theory, 6(3), 191–202.
Lonsdorf, E. V., Eberly, L. E., & Pusey, A. E. (2004). Sex differences in learning in
chimpanzees. Nature, 428(6984), 715–716.
Mann, J., & Patterson, E. M. (2013). Tool use by aquatic animals. Philosophical Transactions of the Royal Society., 368, 20120424.
Mann, J., Sargeant, B. L., Watson-Capps, J., Gibson, Q., Heithaus, M. R., Connor, R.
C., & Patterson, E. (2008). Why do dolphins carry sponges? PLoS One., 3(12), e3868.
Mann, J., Stanton, M. A., Patterson, E. M., Bienenstock, E. J., & Singh, L. O. (2012).
Social networks reveal cultural behaviour in tool-using dolphins. Nature Communications, 3, 980.
Mastrandrea, R., Squartini, T., Fagiolo, G., & Garlaschelli, D. (2013). Enhanced network reconstruction from irreducible local information. arXiv preprint arXiv:
1307.2104.
McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily
in social networks. Annual Review of Sociology, 27, 415–444.

Culture, Diffusion, and Networks in Social Animals

13

Newman, M. E. J. (2004). Analysis of weighted networks. Physics Review E., 70(5),
056131.
Newman, M. E. J. (2006). Modularity and community structure in networks. Proceedings of the National Academy of Sciences, 103(22), 8577–8582.
Newman, M. E. J. (2010). Networks: An introduction. New York, NY: Oxford University
Press, Inc.
Opsahl, T., & Panzarasa, P. (2009). Clustering in weighted networks. Social Networks,
31(2), 155–163.
Pachucki, M. A., & Breiger, R. L. (2010). Cultural holes: Beyond relationality in social
networks and culture. Annual Review of Sociology, 36, 205–224.
Palla, G., Barabasi, A.-L., & Vicsek, T. (2007). Quantifying social group evolution.
Nature, 446(7136), 664–667.
Palla, G., Derenyi, I., Farkas, I., & Vicsek, T. (2005). Uncovering the overlapping community structure of complex networks in nature and society. Nature, 435, 814.
Patterson, E. M., & Mann, J. (2011). The ecological conditions that favor tool use and
innovation in wild bottlenose dolphins (Tursiops sp.). PloS One, 6(7), e22243.
Pinter-Wollman, N., Hobson, E.A. Smith, J.E., Edelman, A.J. Shizuka, D., de Silva, S.,
… , McDonald, D.B. (2014). The dynamics of animal social networks: Analytical,
conceptual, and theoretical advances. Behavioral Ecology, 25(2), 242–255.
Rankin, R., Mann, J., Singh, L. O., Patterson, E. M., Krzyszczyk, E. B., & Bejder, L. (in
review). Dolphin social structure driven by weighted and topological information:
A null model approach. Animal Behaviour.
Rendell, L., Fogarty, L., & Laland, K. N. (2011). Runaway cultural niche construction. Philosophical Transactions of the Royal Society B: Biological Sciences, 366(1566),
823–835.
Rendell, L. E., & Whitehead, H. (2003). Vocal clans in sperm whales (Physeter macrocephalus). Proceedings of the Royal Society of London. Series B: Biological Sciences,
270(1512), 225–231.
Sargeant, B. L., & Mann, J. (2009). Developmental evidence for foraging traditions in
wild bottlenose dolphins. Animal Behaviour, 78(3), 715–721.
Shannon, G., Slotow, R., Durant, S. M., Sayialel, K. N., Poole, J., Moss, C., & McComb,
K. (2013). Effects of social disruption in elephants persist decades after culling.
Frontiers in Zoology, 10(1), 62.
Sharara, H., Singh, L., Getoor, L., & Mann, J. (2011). Understanding actor loyalty to
event-based groups in affiliation networks. Social Network Analysis and Mining, 1,
115–126.
Sharara, H., Singh, L., Getoor, L., & Mann, J. (2012). Finding prominent actors in
dynamic affiliation networks. Human Journal, 1(1), 1–14.
Shen, H., Cheng, X., Cai, K., & Hu, M.-B. (2009). Detect overlapping and hierarchical
community structure in networks. Physica A: Statistical Mechanics and its Applications., 388(8), 1706–1712.
Singh, L. O., Bienenstock, E. J., & Mann, J. (2010). What are we missing? Perspectives on social network analysis for observational scientific data. In B. Furht (Ed.),
Handbook of social networks: Technologies and applications (pp. 147–168). New York,
NY: Springer Science & Business Media, LLC, Chapter 7.

14

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Snijders, T. A. B. (2002). Markov chain Monte Carlo estimation of exponential random graph models. Journal of Social Structure, 3, 2.
Snijders, T. A. B., Pattison, P., Robins, G. L., & Handcock, M. (2006). New specifications for exponential random graph models. Sociological Methodology, 36(1),
99–153.
Tantipathananandh, C., Berger-Wolf, T., & Kempe, D. (2007). A framework for community identification in dynamic social networks. International Conference for
Knowledge Discovery and Database (KDD), pp. 717–726.
Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications
(Vol. 24). Cambridge, England: Cambridge University Press.
Waters, J. S., & Fewell, J. H. (2012). Information processing in social insect networks.
PLoS One, 7, e40337.
Whitehead, H. (2008). Analyzing animal societies: Quantitative methods for vertebrate
social analysis. Chicago, IL: University of Chicago Press.
Whitehead, H., & Lusseau, D. (2012). Animal social networks as substrate for cultural
behavioural diversity. Journal of Theoretical Biology, 294, 19–28.
Whiten, A., Goodall, J., McGrew, W. C., Nishida, T., Reynolds, V., Sugiyama, Y., … ,
Boesch, C. (1999). Cultures in chimpanzees. Nature. 399(6737), 682–685.
Williams, R., & Lusseau, D. (2006). A killer whale social network is vulnerable to
targeted removals. Biology Letters, 2(4), 497–500.
Yurk, H., Barrett-Lennard, L., Ford, J. K. B., & Matkin, C. O. (2002). Cultural transmission within maternal lineages: Vocal clans in resident killer whales in southern
Alaska. Animal Behaviour, 63(6), 1103–1119.

FURTHER READING
Cantor M, Whitehead H. (2013). The interplay between social networks and culture: Theoretically and among whales and dolphins. Philosophical Transactions of
the Royal Society B 368: 20120340. http://dx.doi.org/10.1098/rstb.2012.0340
Hoppitt, W., Boogert, N. J., & Laland, K. N. (2010). Detecting social transmission in
networks. Journal of Theoretical Biology, 263(4), 544–555.
Laland, K. N., & Galef, B. G. (Eds.) (2009). The question of animal culture. Cambridge,
MA: Harvard University Press.
Pinter-Wollman, N., E. A. Hobson, J. E. Smith, A. J. Edelman, D. Shizuka, S. de
Silva, … , McDonald, D. B. (2014). The dynamics of animal social networks: Analytical, conceptual, and theoretical advances Behavioral Ecology, 25(2), 242–255.
doi:10.1093/beheco/art047

JANET MANN SHORT BIOGRAPHY
Janet Mann, Professor of Biology and Psychology and Vice Provost for
Research at Georgetown University, earned her PhD at The University of
Michigan with expertise is in the field of animal behavior. Since 1988 her

Culture, Diffusion, and Networks in Social Animals

15

work has focused on social networks, female reproduction, calf development, life history, conservation, tool-use, social learning and culture among
bottlenose dolphins in Shark Bay, Australia. Her long-term study “The Shark
Bay Dolphin Research Project,” tracks over 1600 dolphins throughout their
lives. Mann has published over 80 scientific papers in journals such as Nature
Communications, Philosophical Transactions of the Royal Society, Proceedings of
the National Academy of Sciences, Proceedings of the Royal Society, Biological
Conservation, and Animal Behaviour and in books such as The Question Animal
Culture, The Biology of Traditions, Rational Animals, and Primates and Cetaceans:
Field Research and Conservation of Complex Mammalian Societies. Her edited
volume, Cetacean Societies (University of Chicago Press, 2000), received
several awards. Twice she was a fellow at The Center for Advanced Study in the
Behavioral Sciences at Stanford University. Dr. Mann’s research has received
considerable media attention worldwide, including a BBC Documentary
“The Dolphins of Shark Bay” focusing on her work in 2011. In 2013, Pamela
Turner published a children’s book “The Dolphins of Shark Bay” (Houghton
Mifflin) about Dr. Mann’s research.
http://explore.georgetown.edu/people/mannj2/

LISA SINGH SHORT BIOGRAPHY
Lisa Singh, Associate Professor in Computer Science at Georgetown
University, is an expert in large-scale data mining. She received her PhD
from Northwestern University in 1999. Her research interests include:
mining social networks, data science and analytics, privacy preserving
data mining, anomaly detection, graph databases, and sampling and bias
in social networks. Her research is supported by the National Science
Foundation and the Office of Naval Research. Dr. Singh has worked
extensively with animal data sets and social media data sets. She has
collaborated with researchers across disciplines at Georgetown University (biology, anthropology, medicine, linguistics, foreign serve, etc.), as
well as the University of Maryland, the University of California—Santa
Cruz, Hewlett Packard, the Census Bureau, and Oak Ridge National
Labs. Dr. Singh also serves on organizing and program committees
of the major data mining and database conferences, including KDD,
ICDM, SIGMOD, PVLDB, and ICDE. She is also heavily involved in
initiatives involving women in computer science and computer science
in K-12 education. More information about her work can be found at:
http://cs.georgetown.edu/∼singh.

16

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

RELATED ESSAYS
Mediation in International Conflicts (Political Science), Kyle Beardsley and
Nathan Danneman
Kin-Directed Behavior in Primates (Anthropology), Carol M. Berman
The Sexual Division of Labor (Anthropology), Rebecca Bliege Bird and Brian
F. Codding
Culture and Cognition (Sociology), Karen A. Cerulo
Adaptation for Culture (Anthropology), Jill M. Church
Sexual Behavior (Anthropology), Melissa Emery Thompson
Culture and Social Networks (Sociology), Jan A. Fuhse
Food Sharing (Anthropology), Michael Gurven and Adrian V. Jaeggi
Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie
Emerging Trends in Social Network Analysis of Terrorism and Counterterrorism (Sociology), David Knoke
Reconciliation and Peace-Making: Insights from Studies on Nonhuman
Animals (Anthropology), Sonja E. Koski
Cooperative Breeding and Human Evolution (Anthropology), Karen L.
Kramer
Implications for Human Sciences (Anthropology), Kevin N. Laland and
Michael O’Brien
Cultural Conflict (Sociology), Ian Mullins
Gestural Communication in Nonhuman Species (Anthropology), Simone Pika
Economics and Culture (Economics), Gerard Roland
Production of Culture (Sociology), Vaughn Schmutz and Candace N. Miller
Vocal Communication in Primates (Anthropology), Katie E. Slocombe
Primate Allomaternal Care (Anthropology), Stacey Tecot and Andrea Baden


Culture, Diffusion, and Networks
in Social Animals
JANET MANN and LISA SINGH

Abstract
Long-term studies of social animals provide detailed data on individual attributes,
behaviors, and associations that enable us to explore cultural diffusion in networks.
In this essay, we describe how network science can be used to improve our understanding of cultural and information transmission. After presenting an operational
definition of culture, we discuss methodologies and research questions applicable
to unweighted, weighted, and dynamic networks. We then highlight relevant studies and methods for both descriptive and predictive analyses that have been used to
identify culture and social learning in animal networks. Applying and extending the
techniques presented will improve our understanding of information transmission,
social learning, and embedded subcultures in the context of human networks.

INTRODUCTION
Our survival, success, and ability to exploit resources depend on cumulative
culture, a ubiquitous feature of human societies. Virtually every facet of our
current state was shaped by cultures past; we excel in niche construction,
perhaps to a fault (Laland & O’Brien, 2011; Rendell, Fogarty, & Laland,
2011). Cultural processes also shape nonhuman animal phenotypes, albeit
to a lesser extent than in humans. Nevertheless, animal societies enable us
to study the underlying network properties and processes that are rarely
accessible in human research and investigate the relationship between
these properties and cultural transmission. For example, long-term studies
of social mammals provide multifaceted connections (e.g., interactions,
associations, kinship, location/home range, communication) and individual
properties (i.e., genotypes and phenotypes) that only a handful of human
studies, usually traditional forager societies (e.g., Hadza foragers, Apicella,
Marlowe, Fowler, & Christakis, 2012) measure. Although we cannot interview animals, privacy laws do not protect them from frequent monitoring
such that real-time behavioral data are often available. This level of detail
Emerging Trends in the Social and Behavioral Sciences. Edited by Robert Scott and Stephen Kosslyn.
© 2015 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

allows us to explore the basic properties of cultural diffusion in networks.
Here we examine how the application of network science to social animals
informs our understanding of culture and information transmission. We
highlight relevant studies and methods and then discuss future directions for
those studying both human and animal networks. These efforts complement
those of social scientists (e.g., see Pachucki & Breiger, 2010) in identifying
theoretical and methodological approaches to network science and culture.
Before continuing, a working definition of culture which is applicable or
measurable across species is needed. In a recent influential book, Laland and
Galef invited social scientists and biologists to discuss The Question of Animal
Culture (Laland & Galef, 2009). Although definitions are fiercely contested, all
contributors agreed on two underlying properties of culture. First, the transmission process involves social learning (learning from the actions or products of others) and second, the socially learned behavior must distinguish
between groups (Laland, J. R. Kendal, & R. L. Kendal, 2009). This minimalist
definition generally works in describing animal cultures, but the challenge
of demonstrating social learning in nonexperimental settings remains.
Owing to this challenge, a number of scientists have tried to eliminate ecological and genetic explanations of behavioral differences between groups
as a way to identify social learning and hence leave “culture” as the only
remaining explanation (e.g., Krützen et al., 2005; Whiten et al., 1999). This
’elimination’ method is clearly flawed, since most social phenomena have a
combination of ecological, genetic, and epigenetic components that interact
with social factors (Kappeler, Barrett, Blumstein, & Clutton-Brock, 2013;
Laland & O’Brien, 2011) and one can never prove the null (Laland & Janik,
2006; Sargeant & Mann, 2009). For example, most socially learned traits
that have been deemed cultural in animals involve foraging (e.g., pine-cone
stripping rats, termite fishing chimpanzees, sponging dolphins), but all of
these depend not only on specific ecological conditions, but also on close kin
(typically the mother) and necessarily includes association, maternal effects,
and biased learning from kin (Aisner & Terkel, 1992; Lonsdorf, Eberly, &
Pusey, 2004; Mann et al., 2008; Mann, Stanton, Patterson, Bienenstock, &
Singh, 2012). To date, few would doubt that social, ecological, demographic,
and genetic factors interact to shape animal social networks and cultural
phenomena embedded in those networks. This multitude of intrinsic and
extrinsic factors receives less focus in human studies, possibly because we
tend to attribute social choice to human networks and biological factors to
animal networks. Still, demonstrating social learning among wild animals
is difficult. As a consequence, researchers have focused on developmental
patterns of a behavior and behavior of associates (e.g., Sargeant & Mann,
2009) or used diffusion models in networks (e.g., Franz & Nunn, 2009,
Hoppitt, Boogert, & Laland, 2010) to measure social transmission.

Culture, Diffusion, and Networks in Social Animals

3

In the last decade, social network studies in the field of animal behavior
have accelerated. For example, in three of the mainstream journals, Animal
Behaviour, Behavioral Ecology, and Behavioral Ecology and Sociobiology, there
were no network studies in 2004 or 2005, one in 2006 and by 2009, 15–21
articles were published cumulatively per year (Science Citation Index search
with keyword “social network.*” This trend has continued. Along with the
increase in animal network research, a plethora of studies began focusing
on behavioral traditions and animal culture, with the specific goal of defining culture by its social transmission properties (i.e., social learning), which
naturally led to defining the underlying properties of culture using social
network analysis.
CULTURAL ANALYSIS USING SOCIAL NETWORKS
Network science is an emerging discipline that studies network representations and predictive models as a way to both explain and predict various
physical, social and biological phenomena (Easley & Kleinberg, 2010; Newman, 2010). In cultural analysis, networks are advantageous for investigating
questions at different scales from the individual (ego networks) to groups
and the network structure as a whole, where the size of the network may
range from a few to billions of individuals. Network analysis and graph theory can be used to help explain the connection between the functionality of a
group and the behavior of different members of the group (Pinter-Wollman
et al., 2014). Further, patterns of information flow both depend on network
structure and influence network structure. Unraveling this relationship is
necessary to understand the relationship between information dissemination
and social learning, that is, cultural processes. However, network structure is
not equivalent to social transmission. To understand those processes, behavioral sampling of individuals in the network is needed. This is an area where
behavioral ecologists excel.
At the basic level, networks are just a collection of points (typically referred
to as nodes, actors, or vertices) connected by lines (typically referred to as edges,
ties, links, or arcs). For simple analyses, we may consider only a simple network in which the nodes are all the same type, for example, people, animals,
organizations, proteins, or computer systems, and the edges connect two
nodes based on a relationship between the two nodes. Example relationships
include kinship, friendship, alliance partner, professional affiliation, and
email correspondence. Social network analysis allows for multiple granularities of analysis and can be beneficial for answering macro-, meso- and
micro-level questions. Examples of the macro-level questions might concern
network density, the number of individuals and paths in the network,
and the distribution of connections. Connectivity can follow a range of

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

distributions, such as random, small-world (high clustering), regular lattice
(no clustering, low heterogeneity, low randomness, and high path lengths),
or scale-free (moderate heterogeneity and randomness—many small world
networks are also scale-free). Meso-level features include distinctiveness of
clusters, community composition, centrality or isolation of communities,
and whether local neighborhoods are tightly connected. At the micro-level,
we might be interested in identifying the information brokers, hubs or highly
connected individuals or isolates. Answering such questions can inform
descriptive and predictive models on cultural processes across micro, meso,
and macro network structures. Still, node and edge attributes (i.e., cultural
behaviors) are needed to identify, quantify and model social transmission.
In a simple network model, the edges do not show the direction of the relationship, the type of relationship or the strength of the relationship. Depending on the analysis, adding one or more of these features can improve the
depth of the analysis and remove potential bias (Singh, Bienenstock, & Mann,
2010). For example, the strength of a relationship can be shown in a network
by adding weights to each edge (Wasserman & Faust, 1994). Generally, for
social networks, weights are values between zero and one. However, negative weights can be used to represent different levels of animosity between
individuals (Newman, 2004). Weighted networks inform on strong and weak
relationships and communities, including channels of high information flow,
that is, likely paths for social information transmission. Finally, adding direction to relationships enables researchers to pose questions related to relationship reciprocity and dominance (Carrington, Scott, & Wasserman, 2005). As
social systems become more complex, network analysis becomes more useful
because of its ability to accommodate features of social complexity such as
motif analysis, hierarchies, individual recognition, and the exponential “cognitive load” faced with an increasing number of social relationships (e.g.,
Dunbar, 2012).
Figure 1 shows a small example of two networks, a simple unweighted,
binary, uni-mode on the left, and a richer weighted, directed, uni-mode network on the right. Colors are used to show clusters in the networks. The
unweighted network is sparsely connected (reducing the possible number of
paths for information flow) and has two clusters with a single edge (in red)
between the clusters. Even though this network is simple, we can still see
that the composition of the two clusters is different. The blue one has a central individual that controls information flow, while the yellow one contains
a clique within it, potentially allowing for more rapid flow of information.
Because there is only one edge between the two clusters, the potential for
information flow between clusters is reduced. The weighted, directed network is also sparse. However, because of the directionality of the edges, we

Culture, Diffusion, and Networks in Social Animals

5

0.4
0.6
0.9

0.8

0.3

0.1
0.7
0.7
(a)

0.9

0.1

0.8

0.3
0.7

(b)

Figure 1 Example networks: (a) unweighted and undirected; (b) weighted and
directed.

can see that information flows from the orange nodes to the green and purple. In addition, as weights are used to capture relationship strength, we also
see that there is a mix of strong and weak relationships through the network,
and while information flow is possible, a message or a piece of information
is unlikely to travel to all the nodes. While both types of networks are informative, relationship strength and reciprocity are important factors in cultural
diffusion.
In a recent paper, Pinter-Wollman et al. (2014) provoked behavioral ecologists to think about moving beyond descriptive analyses of observed patterns, to testing specific hypotheses and predictions regarding the function
of network structures. For example, even though patterns of behavior might
correlate with associations in a network, suggestive of social learning, that
does not explain what drives the behavior or the association. Most literature,
to date has focused on descriptive analysis because of the limited number
of techniques available for predictive analysis, particularly in the context of
more complex, dynamic networks. We now highlight descriptive and predictive approaches used in the literature for identifying and modeling structures
and groups in these different types of networks.
DESCRIPTIVE SOCIAL NETWORK ANALYSIS
Presumably, social transmission predominates in local or embedded communities in a network. Literature from physics and computer science focuses
on measures of cohesion and clustering to identify communities or subsets
of individuals that are more densely connected to each other than expected
(Girvan & Newman, 2002; Newman, 2006; Palla, Barabasi, & Vicsek, 2005,
2007, Shen, Cheng, Cai, & Hu, 2009). Different measures are used to identify
communities. For example, Girvan and Newman (2002) remove edges with
high betweeness (the fraction of shortest paths that traverse an edge) to

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

identify communities. Another approach proposed by Newman (2006)
partitions a network into one with high modularity (the fraction of edges in
a group minus the expected fraction if the network was random) to identify
communities. While these two methods propose algorithms that identify
nonoverlapping communities, Palla et al. (2005) propose using a clique
percolation method that finds maximal cliques to identify overlapping
communities. If social transmission is taking place, then these communities
that are based on network topology would also exhibit similar behavior
(potential subcultures).
While these works focused on static, binary networks, recent techniques have begun considering how communities change and evolve
over time (Backstrom, Huttenlocher, Kleinberg, & Lan, 2006; Caravelli,
Wei, Subak, Singh, & Mann, 2013; Gorke, Hartmann, & Wagner, 2009;
Tantipathananandh, Berger-Wolf, & Kempe, 2007) and how users behave
in these groups (Sharara, Singh, Getoor, & Mann, 2011, 2012). Looking at
these dynamic groups in the context of social transmission, we can measure
if “self-selection” is taking place and individuals are attracted to each other
based on socially learned traits. That is, modularity has a reciprocal nature
in networks, increasing cohesion and social transmission at the same time.
In animal networks, killer whales exhibit similar “dialects” and calls within
matrilineal units (subcommunities), clearly via shared association with kin
(e.g., Yurk, Barrett-Lennard, Ford, & Matkin, 2002). Similarly, sperm whale
matrilineal units use distinct codas that also appear to be socially learned
(Rendell & Whitehead, 2003). In both cases, the communication system is
used for cohesion. It is rare in animal societies however, that there is high
behavioral heterogeneity such that subcultural units within a larger network
can be readily identified. Typically, all members of a community engage in
the socially learned behavior (e.g., termite fishing, Whiten et al., 1999).
Community structures have been analyzed to identify probable cases of
social learning and culture in animal societies (Cantor & Whitehead, 2013).
A few have used dynamic approaches to investigate cultural transmission in
animal networks, such as lobtail foraging in humpback whales in the North
Atlantic (Allen, Weinrich, Hoppitt, & Rendell, 2013). In this case, they used
an order of acquisition analysis (see Hoppitt et al., 2010) to examine diffusion
in humpback whale networks over time. The detail on dynamic interactions between naïve and knowledgeable individuals was weak, although
the pattern over decades was strongly suggestive of social transmission.
Dynamic approaches are particularly valuable for investigation of group
structure evolution and the changing dynamics of group membership. For
example, Caravelli et al. (2013) adjust static community detection algorithms
to dynamic ones to better understand the evolution of communities over
time. The authors also develop metrics related to frequency of appearance of

Culture, Diffusion, and Networks in Social Animals

7

individuals in groups over time to better understand the longevity of social
relationships. Dynamic measures such as stability and diversity in group
participation, where stable actors are those who participate in the same
group over time, while diverse actors are those who consistently participate
across a number of different groups over time (Sharara et al., 2012) can also
serve as a tool for understanding cultural change and stability. A variety of
studies (e.g., Allen et al., 2013; Blonder, Wey, Dornhaus, James, & Sih, 2012;
Boogert, Reader, Hoppitt, & Laland, 2008) have used dynamic methods for
unveiling social transmission in networks.
Binary networks have received far more attention in human networks than
animal networks, possibly because of the view that weighted networks provide similar information as binary in terms of topology (Garlaschelli & Loffredo, 2009; Mastrandrea, Squartini, Fagiolo, & Garlaschelli, 2013; although
see Rankin et al. submitted), but also because, except in social media and
phone networks, we rarely have weighted information in human networks.
Behavioral ecologists typically collect weighted data on their subjects such
as time together or rates of interaction. Such weights are considered critical
components of information transmission (e.g., Whitehead & Lusseau, 2012)
and are presumably relevant in human societies where social relationships
span a continuum based on such factors as frequency, closeness/intimacy,
strength, importance, and valence.
A common approach for computing weights in animal social networks is
the social affinity or association indices (Whitehead, 2008). These measures
account for the number of times each individual is “sighted” alone and with
every other individual to create a ratio for each pair of individuals ranging
from 0 to 1 where 1 indicates that the pair is always together. The strength
of social affinity is that it is an asymmetric weight that maintains relationship direction, capturing individual’s relative sociability and sighting rate
independently of other individuals in the network. In other literatures, traditional community detection algorithms are adjusted to consider weights
(Newman, 2004; Opsahl & Panzarasa, 2009). For example, Newman calculates the betweeness of edges as if weights do not exist, and then divides
the betweenness by the weight of the edge before partitioning the network
into communities. Opsahl and Panzarasa propose using a generalized global
clustering coefficient as a measure to identify members of the same community. The strength of such methods depends on how the weights are initially
computed.
Another direction considers identifying key individuals involved in information transmission processes. Several studies have identified key individuals in information transmission. Some of these approaches involve actual
or modeled targeted removals to determine how information flow might

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

be disrupted or other social changes take place. In the study by Flack, Girvan, de Waal, and Krakauer (2006), removal of specific pigtail macaques that
served as “peacemakers” or “police-monkeys” in captive groups disrupted
the social structure and would presumably impact social transmission. In a
different approach, Williams and Lusseau (2006) simulated the consequences
of targeted versus random removals of killer whales in a wild population
and demonstrated that targeted removals fragmented social units and would
likely disrupt social transmission. Actual removals from culling or poaching
among African elephants can have social impacts that last for decades largely
because cultural information is lost (Archie & Chiyo, 2012; Shannon et al.,
2013).
At the heart of information transmission is determining how to model the
transmission of ideas. One direction of research investigates ways to model
the flow through a network. Similar to disease transmission, we can model
nodes in a network as active, for example, informed or influenced, or inactive.
Active nodes can then use the edges of the network to spread the contagion (e.g. information). For example, Domingos and Richardson (2001) used a
global, probabilistic model that employed the joint distribution of the behavior over all the nodes to find influential ones. Kempe, Kleinberg, & Tardos
(2003) used a diffusion process that begins with an initial set of active nodes
and used different weighting schemes to determine whether or not a neighbor should be activated.
PREDICTIVE SOCIAL NETWORK ANALYSIS
Traditional statistical methods can be difficult to use with relational data
because by definition, the members of the network are not independent from
each other. In fact, it is the relationship between individuals that is typically
of interest in the analysis. This lack of independence makes it difficult to interpret with traditional statistical tests. However, there are some methods that
are particularly exciting for hypothesis testing about the function and role of
different interaction patterns and relationships.
Relational variables tend to be central for predictive analyses that involve
networks. Two approaches that have been widely used to model dependencies between relational variables are multiple regression quadratic
assignment procedure (MRQAP) (Dekker, Krackhardt, & Snijders, 2007;
Krackhardt, 1988) and exponential random graph models (ERGM) (Snijders,
2002; Snijders, Pattison, Robins, & Handcock, 2006). Instead of variables
that are vectors of values, both methods have variables that are entries
in an adjacency matrix to capture the relational aspect of the data. Both
methods are meant for cross sectional network data analysis, however, they

Culture, Diffusion, and Networks in Social Animals

9

are designed to answer slightly different questions. (The discussion below is
based on the ones presented in Dekker et al., (2007) and Snijders (2002).)
MRQAP was designed to investigate factors affecting pairwise associations.
Given two different square matrices, is there an association between the same
entries in these different matrices? MRQAP is a relational version of a standard regression analysis that can be used on weighted or binary networks.
The procedure itself is a set of permutation tests for multiple linear regression model coefficients. The determinants are at the relationship or dyad level
and the dependencies generated by the network structure as a whole are controlled for. Questions of interest are typically of the form—How do different
factors (age, gender, etc.) influence the strength of association in undirected,
directed, weighted or binary dependent variables?
We recently applied the MRQAP to investigate whether individuals group
because they share a cultural trait—in contrast to animals sharing a cultural
trait because they group (e.g., killer whale dialects, Yurk et al., 2002). We study
sponge tool use in wild bottlenose dolphins where the dolphins (spongers)
use basket shaped marine sponges to ferret prey from the seafloor (Mann
et al., 2008; Patterson & Mann, 2011). Only about 5% of community members
use sponges in this way (Mann & Patterson, 2013), but spongers associate
regularly with nonspongers (Mann et al., 2012). Our recent work demonstrated that a subset of the community preferentially associate based on their
tool-using status, when sex, location, and maternal kinship are controlled for
(Mann et al., 2012). Because dolphins learn “sponging” from their mothers
and tend to be solitary while hunting with their sponge tools (Mann et al.,
2008), we were able to show homophily based on tool-using status, going a
step beyond others have shown to date. That is, spongers appear to prefer to
be with each other, suggesting that they identify with others similar to themselves (Mann et al., 2012). This is similar to patterns found in human social
groups (McPherson, Smith-Lovin, & Cook, 2001). Homophily based on similar age, sex, reproductive state or other factors is common in animals, but
they usually share the cultural trait because they group, not group based on
shared cultural traits (Mann et al., 2012).
In contrast, ERGM is designed to model networks as a whole, considering
dependencies between different relational variables jointly. This allows one
to model structural dependencies in the network. ERGM can still be used
to model dependencies between relational variables while controlling for
network structure, but it is well suited for modeling structural dependencies
of the network. In an ERGM analysis, relationships can be directed or
undirected, but the relationships must be binary. An ERGM analysis uses
stochastic modeling to determine the probability that a connection exists
between individuals based on some set of predictor variables. The explanatory variables may be attributes of the individuals, dyadic, or network

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

features. Questions of interest are typically of the form—How do various
factors influence the structure of binary networks? Henrich and Broesch
(2011) used ERGM to study transmission processes of critical cultural
information on medicinal plants, fishing, and yam farming among Fijian
villagers and had fascinating results. They found, for example, that social
learning biases were stronger towards those who demonstrated success at
a given behavior (fishing and farming) than knowledge per se (Henrich &
Broesch, 2011). ERGM is used extensively in the social sciences, but has
recently been applied to animal networks although not in the domain of
culture (Pinter-Wollman et al., 2014; Ilany, Barocas, Koren, Kam, & Geffen,
2013). One limitation with ERGM is that weighted approaches have only
recently been developed (Krivitsky, 2012).
SUMMARY AND NEW DIRECTIONS
As more people share behavioral information online publicly, researchers
will have the opportunity to better understand human behavior and the
influence of social relationships on this behavior. This essay presented
different studies and methods that have been proposed for identifying
culture in animal societies. Applying and extending both the descriptive
and predictive technique presented will improve our understanding of
information transmission and social learning in the context of human
networks. This in turn may help researchers identify subcultures that are
embedded in human networks.
As mentioned previously, many of the methods for descriptive analysis do
not consider complex networks that contain multiple node types, multiple
edge types, weights, reciprocity, and attributes. Including all of these network features is important for more complete descriptive analysis. Similar
extensions are needed for inference models as well. Without them, our ability
to answer questions related to culture and social learning will remain limited.
Acknowledgments: We are grateful to our colleagues on The Shark Bay
Dolphin Research Project and our funding sources: NSF, 0941487, 0918308,
ONR 10230702, the National Geographic Society Committee for Research and
Exploration and Georgetown University.
REFERENCES
Aisner, R., & Terkel, J. (1992). Ontogeny of pine cone opening behaviour in the black
rat, Rattus rattus. Animal Behaviour, 44, 327–336.
Allen, J., Weinrich, M., Hoppitt, W., & Rendell, L. (2013). Network-based diffusion
analysis reveals cultural transmission of lobtail feeding in humpback whales. Science, 340(6131), 485–488.

Culture, Diffusion, and Networks in Social Animals

11

Apicella, C. L., Marlowe, F. W., Fowler, J. H., & Christakis, N. A. (2012). Social networks and cooperation in hunter-gatherers. Nature, 481(7382), 497–501.
Archie, E. A., & Chiyo, P. I. (2012). Elephant behaviour and conservation: Social relationships, the effects of poaching, and genetic tools for management. Molecular
Ecology, 21(3), 765–778.
Bacher, K., Allen, S., Lindholm, A. K., Bejder, L., & Krützen, M. (2010). Genes or
culture: Are mitochondrial genes associated with tool use in bottlenose dolphins
(Tursiops sp.)? Behavior Genetics, 40(5), 706–714.
Backstrom, L., Huttenlocher, D., Kleinberg, J., & Lan, X. (2006). Group formation in
large social networks: Membership, growth, and evolution. International Conference for Knowledge Discovery and Database (KDD), pp. 44–54.
Blonder, B., Wey, T., Dornhaus, A., James, R., & Sih, A. (2012). Temporal dynamics
and network analysis. Methods in Ecology and Evolution, 3, 958–972.
Boogert, N. J., Reader, S. M., Hoppitt, W., & Laland, K. N. (2008). The origin and
spread of innovations in starlings. Animal Behaviour, 75(4), 1509–1518.
Cantor, M., & Whitehead, H. (2013). The interplay between social networks and culture: Theoretically and among whales and dolphins. Philosophical Transactions of
the Royal Society B, 368, 20120340.
Caravelli, P., Wei, Y., Subak, D., Singh, L., & Mann, J. (2013). Understanding evolving
group structures in time-varying network. Advances in Social Networks Analysis and
Mining, 142–148.
Carrington, P., Scott, J., & Wasserman, S. (2005). Models and methods in social network
analysis. New York, NY: Cambridge University Press.
Dekker, D., Krackhardt, D., & Snijders, T. A. (2007). Sensitivity of MRQAP tests to
collinearity and autocorrelation conditions. Psychometrika, 72, 563–581.
Domingos, P. & Richardson, M. (2001). Mining the network value of customers. ACM
International Conference on Knowledge Discovery and Data Mining.
Dunbar, R. I. (2012). The social brain meets neuroimaging. Trends in Cognitive Sciences,
16(2), 101–102.
Easley, D., & Kleinberg, J. (2010). Networks, crowds, and markets: Reasoning about a
highly connected world. New York, NY: Cambridge University Press.
Flack, J. C., Girvan, M., de Waal, F. B., & Krakauer, D. C. (2006). Policing stabilizes
construction of social niches in primates. Nature, 439, 426–429.
Franz, M., & Nunn, C. L. (2009). Network-based diffusion analysis: a new method
for detecting social learning. Proceedings of the Royal Society B: Biological Sciences,
276, 1829–1836.
Garlaschelli, D., & Loffredo, M. I. (2009). Generalized bose-fermi statistics and
structural correlations in weighted networks. Physical Review Letters, 102(3),
038701.
Girvan, M., & Newman, M. E. J. (2002). Community structure in social and biological
networks. Proceedings of the National Academy of Sciences, 99(2), 7821–7826.
Gorke, R., Hartmann, T., & Wagner, D. (2009). Dynamic graph clustering using
minimum-cut trees. Proceedings of the 11th International Symposium on Algorithms and Data Structures (WADS ’09), pp. 339–350.

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Henrich, J., & Broesch, J. (2011). On the nature of cultural transmission networks: Evidence from Fijian villages for adaptive learning biases. Philosophical Transactions of
the Royal Society B: Biological Sciences, 366(1567), 1139–1148.
Hoppitt, W., Boogert, N. J., & Laland, K. N. (2010). Detecting social transmission in
networks. Journal of Theoretical Biology, 263(4), 544–555.
Ilany, A., Barocas, A., Koren, L., Kam, M., & Geffen, E. (2013). Structural balance in
the social networks of a wild mammal. Animal Behaviour, 85(6), 1397–1405.
Kappeler, P. M., Barrett, L., Blumstein, D. T., & Clutton-Brock, T. H. (2013). Constraints and flexibility in mammalian social behaviour: Introduction and synthesis. Philosophical Transactions of the Royal Society B: Biological Sciences, 368(1618),
20120337.
Kempe, D., Kleinberg, J., & Tardos, E. (2003). Maximizing the spread of influence
through a social network. ACM International Conference on Knowledge Discovery and Data Mining, pp. 137–146.
Krackhardt, D. (1988). Predicting with networks—Nonparametric multipleregression analysis of dyadic data. Social Networks, 10(4), 359–381.
Krivitsky, P. N. (2012). Exponential-family random graph models for valued networks. Electronic Journal of Statistics, 6, 1100–1128.
Krützen, M., Mann, J., Heithaus, M. R., Connor, R. C., Bejder, L., & Sherwin, W. B.
(2005). Cultural transmission of tool use in bottlenose dolphins. Proceedings of the
National Academy of Sciences of the United States of America, 102(25), 8939–8943.
Laland, K. N., & Galef, B. G. (Eds.) (2009). The question of animal culture. Cambridge,
MA: Harvard University Press.
Laland, K. N., & Janik, V. M. (2006). The animal cultures debate. Trends in Ecology &
Evolution, 21, 542.
Laland, K. N., Kendal, J. R., & Kendal, R. L. (2009). Animal culture: Problems and
solutions. In K. N. Laland & B. G. Galef Jr., (Eds.), The question of animal culture
(pp. 174–197). Cambridge, MA: Harvard University Press.
Laland, K. N., & O’Brien, M. J. (2011). Cultural niche construction: An introduction.
Biological Theory, 6(3), 191–202.
Lonsdorf, E. V., Eberly, L. E., & Pusey, A. E. (2004). Sex differences in learning in
chimpanzees. Nature, 428(6984), 715–716.
Mann, J., & Patterson, E. M. (2013). Tool use by aquatic animals. Philosophical Transactions of the Royal Society., 368, 20120424.
Mann, J., Sargeant, B. L., Watson-Capps, J., Gibson, Q., Heithaus, M. R., Connor, R.
C., & Patterson, E. (2008). Why do dolphins carry sponges? PLoS One., 3(12), e3868.
Mann, J., Stanton, M. A., Patterson, E. M., Bienenstock, E. J., & Singh, L. O. (2012).
Social networks reveal cultural behaviour in tool-using dolphins. Nature Communications, 3, 980.
Mastrandrea, R., Squartini, T., Fagiolo, G., & Garlaschelli, D. (2013). Enhanced network reconstruction from irreducible local information. arXiv preprint arXiv:
1307.2104.
McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily
in social networks. Annual Review of Sociology, 27, 415–444.

Culture, Diffusion, and Networks in Social Animals

13

Newman, M. E. J. (2004). Analysis of weighted networks. Physics Review E., 70(5),
056131.
Newman, M. E. J. (2006). Modularity and community structure in networks. Proceedings of the National Academy of Sciences, 103(22), 8577–8582.
Newman, M. E. J. (2010). Networks: An introduction. New York, NY: Oxford University
Press, Inc.
Opsahl, T., & Panzarasa, P. (2009). Clustering in weighted networks. Social Networks,
31(2), 155–163.
Pachucki, M. A., & Breiger, R. L. (2010). Cultural holes: Beyond relationality in social
networks and culture. Annual Review of Sociology, 36, 205–224.
Palla, G., Barabasi, A.-L., & Vicsek, T. (2007). Quantifying social group evolution.
Nature, 446(7136), 664–667.
Palla, G., Derenyi, I., Farkas, I., & Vicsek, T. (2005). Uncovering the overlapping community structure of complex networks in nature and society. Nature, 435, 814.
Patterson, E. M., & Mann, J. (2011). The ecological conditions that favor tool use and
innovation in wild bottlenose dolphins (Tursiops sp.). PloS One, 6(7), e22243.
Pinter-Wollman, N., Hobson, E.A. Smith, J.E., Edelman, A.J. Shizuka, D., de Silva, S.,
… , McDonald, D.B. (2014). The dynamics of animal social networks: Analytical,
conceptual, and theoretical advances. Behavioral Ecology, 25(2), 242–255.
Rankin, R., Mann, J., Singh, L. O., Patterson, E. M., Krzyszczyk, E. B., & Bejder, L. (in
review). Dolphin social structure driven by weighted and topological information:
A null model approach. Animal Behaviour.
Rendell, L., Fogarty, L., & Laland, K. N. (2011). Runaway cultural niche construction. Philosophical Transactions of the Royal Society B: Biological Sciences, 366(1566),
823–835.
Rendell, L. E., & Whitehead, H. (2003). Vocal clans in sperm whales (Physeter macrocephalus). Proceedings of the Royal Society of London. Series B: Biological Sciences,
270(1512), 225–231.
Sargeant, B. L., & Mann, J. (2009). Developmental evidence for foraging traditions in
wild bottlenose dolphins. Animal Behaviour, 78(3), 715–721.
Shannon, G., Slotow, R., Durant, S. M., Sayialel, K. N., Poole, J., Moss, C., & McComb,
K. (2013). Effects of social disruption in elephants persist decades after culling.
Frontiers in Zoology, 10(1), 62.
Sharara, H., Singh, L., Getoor, L., & Mann, J. (2011). Understanding actor loyalty to
event-based groups in affiliation networks. Social Network Analysis and Mining, 1,
115–126.
Sharara, H., Singh, L., Getoor, L., & Mann, J. (2012). Finding prominent actors in
dynamic affiliation networks. Human Journal, 1(1), 1–14.
Shen, H., Cheng, X., Cai, K., & Hu, M.-B. (2009). Detect overlapping and hierarchical
community structure in networks. Physica A: Statistical Mechanics and its Applications., 388(8), 1706–1712.
Singh, L. O., Bienenstock, E. J., & Mann, J. (2010). What are we missing? Perspectives on social network analysis for observational scientific data. In B. Furht (Ed.),
Handbook of social networks: Technologies and applications (pp. 147–168). New York,
NY: Springer Science & Business Media, LLC, Chapter 7.

14

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Snijders, T. A. B. (2002). Markov chain Monte Carlo estimation of exponential random graph models. Journal of Social Structure, 3, 2.
Snijders, T. A. B., Pattison, P., Robins, G. L., & Handcock, M. (2006). New specifications for exponential random graph models. Sociological Methodology, 36(1),
99–153.
Tantipathananandh, C., Berger-Wolf, T., & Kempe, D. (2007). A framework for community identification in dynamic social networks. International Conference for
Knowledge Discovery and Database (KDD), pp. 717–726.
Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications
(Vol. 24). Cambridge, England: Cambridge University Press.
Waters, J. S., & Fewell, J. H. (2012). Information processing in social insect networks.
PLoS One, 7, e40337.
Whitehead, H. (2008). Analyzing animal societies: Quantitative methods for vertebrate
social analysis. Chicago, IL: University of Chicago Press.
Whitehead, H., & Lusseau, D. (2012). Animal social networks as substrate for cultural
behavioural diversity. Journal of Theoretical Biology, 294, 19–28.
Whiten, A., Goodall, J., McGrew, W. C., Nishida, T., Reynolds, V., Sugiyama, Y., … ,
Boesch, C. (1999). Cultures in chimpanzees. Nature. 399(6737), 682–685.
Williams, R., & Lusseau, D. (2006). A killer whale social network is vulnerable to
targeted removals. Biology Letters, 2(4), 497–500.
Yurk, H., Barrett-Lennard, L., Ford, J. K. B., & Matkin, C. O. (2002). Cultural transmission within maternal lineages: Vocal clans in resident killer whales in southern
Alaska. Animal Behaviour, 63(6), 1103–1119.

FURTHER READING
Cantor M, Whitehead H. (2013). The interplay between social networks and culture: Theoretically and among whales and dolphins. Philosophical Transactions of
the Royal Society B 368: 20120340. http://dx.doi.org/10.1098/rstb.2012.0340
Hoppitt, W., Boogert, N. J., & Laland, K. N. (2010). Detecting social transmission in
networks. Journal of Theoretical Biology, 263(4), 544–555.
Laland, K. N., & Galef, B. G. (Eds.) (2009). The question of animal culture. Cambridge,
MA: Harvard University Press.
Pinter-Wollman, N., E. A. Hobson, J. E. Smith, A. J. Edelman, D. Shizuka, S. de
Silva, … , McDonald, D. B. (2014). The dynamics of animal social networks: Analytical, conceptual, and theoretical advances Behavioral Ecology, 25(2), 242–255.
doi:10.1093/beheco/art047

JANET MANN SHORT BIOGRAPHY
Janet Mann, Professor of Biology and Psychology and Vice Provost for
Research at Georgetown University, earned her PhD at The University of
Michigan with expertise is in the field of animal behavior. Since 1988 her

Culture, Diffusion, and Networks in Social Animals

15

work has focused on social networks, female reproduction, calf development, life history, conservation, tool-use, social learning and culture among
bottlenose dolphins in Shark Bay, Australia. Her long-term study “The Shark
Bay Dolphin Research Project,” tracks over 1600 dolphins throughout their
lives. Mann has published over 80 scientific papers in journals such as Nature
Communications, Philosophical Transactions of the Royal Society, Proceedings of
the National Academy of Sciences, Proceedings of the Royal Society, Biological
Conservation, and Animal Behaviour and in books such as The Question Animal
Culture, The Biology of Traditions, Rational Animals, and Primates and Cetaceans:
Field Research and Conservation of Complex Mammalian Societies. Her edited
volume, Cetacean Societies (University of Chicago Press, 2000), received
several awards. Twice she was a fellow at The Center for Advanced Study in the
Behavioral Sciences at Stanford University. Dr. Mann’s research has received
considerable media attention worldwide, including a BBC Documentary
“The Dolphins of Shark Bay” focusing on her work in 2011. In 2013, Pamela
Turner published a children’s book “The Dolphins of Shark Bay” (Houghton
Mifflin) about Dr. Mann’s research.
http://explore.georgetown.edu/people/mannj2/

LISA SINGH SHORT BIOGRAPHY
Lisa Singh, Associate Professor in Computer Science at Georgetown
University, is an expert in large-scale data mining. She received her PhD
from Northwestern University in 1999. Her research interests include:
mining social networks, data science and analytics, privacy preserving
data mining, anomaly detection, graph databases, and sampling and bias
in social networks. Her research is supported by the National Science
Foundation and the Office of Naval Research. Dr. Singh has worked
extensively with animal data sets and social media data sets. She has
collaborated with researchers across disciplines at Georgetown University (biology, anthropology, medicine, linguistics, foreign serve, etc.), as
well as the University of Maryland, the University of California—Santa
Cruz, Hewlett Packard, the Census Bureau, and Oak Ridge National
Labs. Dr. Singh also serves on organizing and program committees
of the major data mining and database conferences, including KDD,
ICDM, SIGMOD, PVLDB, and ICDE. She is also heavily involved in
initiatives involving women in computer science and computer science
in K-12 education. More information about her work can be found at:
http://cs.georgetown.edu/∼singh.

16

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

RELATED ESSAYS
Mediation in International Conflicts (Political Science), Kyle Beardsley and
Nathan Danneman
Kin-Directed Behavior in Primates (Anthropology), Carol M. Berman
The Sexual Division of Labor (Anthropology), Rebecca Bliege Bird and Brian
F. Codding
Culture and Cognition (Sociology), Karen A. Cerulo
Adaptation for Culture (Anthropology), Jill M. Church
Sexual Behavior (Anthropology), Melissa Emery Thompson
Culture and Social Networks (Sociology), Jan A. Fuhse
Food Sharing (Anthropology), Michael Gurven and Adrian V. Jaeggi
Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie
Emerging Trends in Social Network Analysis of Terrorism and Counterterrorism (Sociology), David Knoke
Reconciliation and Peace-Making: Insights from Studies on Nonhuman
Animals (Anthropology), Sonja E. Koski
Cooperative Breeding and Human Evolution (Anthropology), Karen L.
Kramer
Implications for Human Sciences (Anthropology), Kevin N. Laland and
Michael O’Brien
Cultural Conflict (Sociology), Ian Mullins
Gestural Communication in Nonhuman Species (Anthropology), Simone Pika
Economics and Culture (Economics), Gerard Roland
Production of Culture (Sociology), Vaughn Schmutz and Candace N. Miller
Vocal Communication in Primates (Anthropology), Katie E. Slocombe
Primate Allomaternal Care (Anthropology), Stacey Tecot and Andrea Baden



Media

Culture, Diffusion, and Networks in Social Animals