ArticlePDF Available

Network measures in animal social network analysis: Their strengths, limits, interpretations and uses

  • University of Strasbourg Institute for Advanced Study


1. We provide an overview of the most commonly used social network measures in animal research for static networks or time-aggregated networks. 2. For each of these measures, we provide clear explanations as to what they measure, we describe their respective variants, we underline the necessity to consider these variants according to the research question addressed, and we indicate considerations that have not been taken so far. 3. We provide a guideline indicating how to use them depending on of the data collection protocol, the social system studied and the research question addressed. Finally, we inform about the existent gaps and remaining challenges in the use of several variants and provide future research directions.
Methods Ecol Evol. 2020;00:1–12.
  1© 2020 British Ecological Society
For those unfamiliar with social network analysis (SNA) terminol-
ogy (highlighted in the text with asterisks), we provide a glossar y in
Table 1. The mathematical formulae of all the network measures dis-
cussed in this manuscript are provided in Appendix S1 and software
handling their computation are in Appendix S2.
Social network analysis has become a methodological framework
that allows a transdisciplinary approach (from proteomic research
to animal societies and ecosystems) to study multiple questions
within single systems (networks*) such as groups, populations as
well as connected units (links* and nodes*) of the systems. For ex-
ample, in the study of animal societies, SNA can reveal the causes
and consequences of individuals’ social heterogeneity (variation in
social behaviour) and link the social interactions to both ecological
and evolutionary processes (Sueur, Romano, Sosa, & Puga-Gonzalez,
2019). Here we describe how the use of specific network measures*
can lead to study these different aspects and levels.
The surge of SNA in the last couple of decades has been ac-
companied by the development of a large number of analytical
software and methods to calculate network measures (Borgatti,
Everett, & Freeman, 20 02; Csardi & Nepusz, 2006; Sosa et al., 2018;
Whitehead, 2009). This has resulted in a diversity of software that
vary in the way some network measures are calculated (because
they used different methods), and/or are specialized in calculations/
functions designed with a specific research purpose, here referred to
as variant s. Not surprisingly, non-exper ts in SNA may find dif ficult to
get a clear picture of the most adequate approaches or tools for their
research question.
In this manuscript, we do not aim to show the usefulness of SNA
(which was already proved many times); instead we provide the
reader with an extensive list of the different measures* (and their
variants) that are commonly used in animal social network analysis
(ASNA) for static networks or time-aggregated networks. We do so
to highlight how mathematical dif ferences in the calculation of these
measurements may af fect the interpretation of results, making it
necessary to indicate some considerations that have not been taken
so far. Our aim was to provide researchers with a guideline that
helps them to: (a) interpret the different measures and their variants,
(b) choose a specific measure according to the research question
and (c) avoid misuses of SNA measures. Although we provide a pre-
scriptive approach on which network measure to use, when and how
Received: 24 September 2019 
  Accepted: 3 January 2020
DOI : 10.1111 /20 41-210X.1336 6
Network measures in animal social network analysis:
Their strengths, limits, interpretations and uses
Sebastian Sosa1| Cédric Sueur1,2 | Ivan Puga-Gonzalez3
1Universi té de Strasbourg , CNRS, IPHC UMR
7178, Strasbourg, France
2Instit ut Univer sitaire de France , Paris,
3Institute for Global Development and
Plannin g, University of A gder, Kristiansand,
Sebastian Sosa
Handling Editor: Timothée Poisot
1. We provide an overview of the most commonly used social network measures in
animal research for static networks or time-aggregated networks.
2. For each of these measures, we provide clear explanations as to what they meas-
ure, we describe their respective variants, we underline the necessity to consider
these variants according to the research question addressed, and we indicate con-
siderations that have not been taken so far.
3. We provide a guideline indicating how to use them depending on the data col-
lection protocol, the social system studied and the research question addressed.
Finally, we inform about the existent gaps and remaining challenges in the use of
several variants and provide future research directions.
animal research, animal social networks, net work measures, social network analysis, social
network measures, static networks, time-aggregated networks
Methods in Ecology and Evoluon
SOSA et Al .
depending on the research question, the data collection protocol
and the species-specific social structure (Figures 1‒3), readers may
keep in mind that SNA is a versatile tool and each research question
and system requires its own, bespoke set of considerations to deal
with its own specificities.
1.1 | Considerations prior to selecting
network measures
Before considering the computation of net work measures, one
may first consider the type of data collected (e.g. rare or frequent,
associations or interactions), the type of system under study
(e.g. cohesive social group, population, etc.), the environment in
which individuals evolve (e.g. forest or open field) and how the
data are collected (e.g. scan sampling, focal sampling, gambit of
the group [GoG]) as each of these fac tors may affect the accu-
racy of the data collected and the extent to which the data are a
fair representation of the system. For example, data collected in
animal social research can be divided into two main categories,
namely associations and interactions. Associations are usually col-
lected with GoG or scan sampling and interactions with scan or
focal samplings. Whereas GoG allows to rapidly collect numer-
ous individual associations, it inevitably generates networks with
Ter m s Definition
Alters Nodes connected to ego
Binary Considering the presence or absence of links between two
Closed triplets Three nodes interconnected between each other
Directed network Network with link directionality (representing the directionalit y
of the behaviour)
Directionality Link directionality from one node to another
Ego A specific node
Ego-network A network with ego's connections only
Incoming links Interaction received
Link Element of a network representing the connection
(e.g. interaction or association) between two nodes. Term edge
is used as synonym in the literature
Micro-motifs Substructures of a network
Network A system of interconnected elements
Network clusterization Formation of subgroups in a network
Network global measures Measures calculated at the level of the whole network
Network measures Mathematical calculations to quantify specific features of a
network, inclu de global, node and polyadic measures
Network node measures Measures calculated at the level of nodes
Network resilience Capacity for the network to remain undisrupted when nodes are
Network transmission
How well pathogens or information spread in the net work
Node Element of a network representing an individual. Term vertice is
used as synonym in the literature
Node centrality A central node is highly connected and/or is connected to highly
connected nodes
Outgoing links Interaction given
Strongest links Links with highest weights
Undirected network Network without link directionality
Unweighted network Networ k in which links represent the presence (1) or absence (0)
of interactions/associations between nodes
Weakest links Links with lowest weights
Weigh t Value of a link usually representing the frequenc y of an
Weighted network Network in which the weights of the links represent the
frequency of interactions/associations between nodes
TABLE 1 Network glossary
Methods in Ecology and Evoluon
SOSA et Al .
higher density than networks based on social interactions that
are generally distributed differently depending on the social part-
ners as well as undirected* networks (Franks, Ruxton, & James,
2010). This aspect entails the following three main considerations:
(a) whether network associations represent faithfully the group/
population social structure, (b) the usefulness of GoG in the study
of social dif fusion such as epidemiology and (c) the use of meas-
ures that do not consider links’ weights* in networks obtained
through GoG. Similarly, the system studied and the environment in
which individuals evolve may make it necessar y to adapt the data
collection protocol. For example, scan sampling can be perfectly
adapted for the study of cohesive species with a well-known
group composition, small size and/or living in an open environ-
ment whereas focal sampling may be preferred for larger cohesive
species living in dense forests or fission-fusion societies (in this
case, scan sampling may lead to oversample the core group easily
visible). As a rule, one may consider that it is not the best choice to
use data obtained through GoG for the study of social diffusion or
the computation of measures that do not consider links’ weights as
this observation protocol produces highly dense networks and the
link filtering usually performed to reduce the density generates
important biases (Franks et al., 2010).
FIGURE 1 Decision tree for examining individual social heterogeneity according to the research question and the network studied
FIGURE 2 Decision tree for examining patterns of individual interactions according to the research question and the network studied
Methods in Ecology and Evoluon
SOSA et Al .
1.2 | Examining heterogeneity in node interactions
Node measures* (Figure 1) enable to assess individuals’ social hetero-
geneity and to understand the underlying mechanisms such as indi-
vidual characteristics (e.g. ageing process; Almeling, Hammerschmidt,
Sennhenn-Reulen, Freund, & Fischer, 2016), ecological factors
(e.g. demographic variation; Borgeaud, Sosa, Sueur, & Bshary, 2017)
and evolutionary processes (e.g. differences in social st yles; Sueur
et al., 2011). Node measures are calculated at an individual level and
assess in d ifferent ways and wi th different mea nings how an indivi dual
is connected. Connections can be ego's* direct links only (e.g. degree,
streng th), alters’* links as well (e.g. eigenvector, clustering coeffi-
cient) or even all the links of the network (e.g. betweenness). Node
measures can also be used to describe the overall network structure
through distributions, means and coefficients of variation.
1.2.1| Degree and strength
The degree measures the number of links of a node. When computed
on an undirected network, the degree represents the number of alters
of ego. When the network is directed*, it represents the number of
either incoming* or outgoing* links of ego and it is then called in-
degree (i.e. number of incoming links) or out-degree (i.e. number of
outgoing links) respectively. In-degree is generally used as a measure
of popularity in affiliative networks and out-degree as a measure of
expansiveness (Borgatti, Everett, & Johnson, 2018). Note that degree
can also be computed in directed networks, in this case it represents
the sum of incoming and outgoing links and not the number of alters.
Strength (or weighted degree) is the sum of links’ weights in a
weighted network*. When the network comprises direc ted links,
then it is also possible to differentiate bet ween in-strength (the sum
of weights of incoming links) and out-strength (the sum of weights
of outgoing links). In ASNA, these measures usually represent the
frequency of individuals’ interactions/associations and thus reflect
individuals’ sociality and social activity. While degree and strength
can be considered correlated, it may not always be the case as indi-
viduals can interact frequently with few social partners or vice versa
(Liao, Sosa, Wu, & Zhang, 2018). Therefore, it is necessary to test
their correlation prior to the analysis.
There is a long list of research that have used degree and
streng th; these are the main findings: Degree has been found to
decrease with age in primates and marmots (Almeling et al., 2016)
while strength does not (Almeling et al., 2016; Liao et al., 2018).
The philopatric sex has shown higher affiliative degree and affil-
iative strength in several species (Borgeaud, et al., 2017) as well
as high-ranked individuals (Brent, Ruiz-Lambides, & Platt, 2017b).
A positive correlation has been found between parasite load and
degree and streng th (Leu, Farine, Wey, Sih, & Bull, 2016), although
this correlation may be compensated by social buffering/suppor t
(Scharf, Modlmeier, Beros, & Foitzik, 2012). Several personality
traits have been positively related to degree and strength such
as exploration (Aplin, Farine, Mann, & Sheldon, 2014) or boldness
(Moyers, Adelman, Farine, Moore, & Hawley, 2018). In several
FIGURE 3 Decision tree for examining group structure and properties according to the research question and the network studied
Methods in Ecology and Evoluon
SOSA et Al .
primate species, the social circle of infants (i.e. mothers’ degrees)
has been found to have a significant impact on their develop-
ment (Shimada & Sueur, 2014). Finally, individuals with wider so-
cial circles show higher longevit y (Brent, Ruiz-Lambides, & Plat t,
2017a; Silk et al., 2010) and greater reproductive success (Schülke,
Bhagavatula, Vigilant, & Ostner, 2010).
Degree shows low sensitivity to observation biases (e.g. misiden-
tification of individuals or unobserved interactions), which makes it
particularly relevant for epidemiology studies (Krause, James, Franks,
& Croft, 2014). However, when considering data collection, due to
the high connectiveness of networks generated by GoG, degree may
be less suitable than strength since degree is strongly correlated with
density. Finally, cautions must be taken when using software as the
computation of degree with directed networks induces by default the
computation of the sum of incoming and outgoing links and not the
number of alters. These contrasting variants of a measure as simple as
the degree serve as a reminder that special care must be taken as to
the mathematical formulae applied to avoid misinterpretations.
1.2.2 | Eigenvector centrality
Eigenvector centrality is the first non-negative eigenvector value
obtained by transforming an adjacency matrix linearly. It can be
computed on weighted, binary*, directed or undirected networks.
It measures the centrality* by examining the connectedness of ego
as well as that of its alters. Thus, a node's eigenvector value can
be linked either to its own degree or strength or to the degrees or
streng ths of the nodes to which it is connected.
Eigenvector may be interpreted as the social support or social
capital of an individual (Brent, Semple, Dubuc, Heistermann, &
MacLarnon, 2011), that is the real or perceived availability of so-
cial resources. Eigenvector has been extensively used in ASNA and
is linked to biological aspects such as individual fitness (Stanton &
Mann, 2012), epidemiology (Balasubramaniam, Beisner, Vandeleest,
Atwill, & McCowan, 2016), individual characteristics (Sosa, 2016) or
social st yle (Sueur et al., 2011).
1.2.3 | Betweenness
Betweenness is the number of times a node is included in the short-
est paths (geodesic distances) generated by every combination of
two nodes. The value of the bet weenness informs on the theoretical
role of a node in the social transmission (information, disease, etc.,
see Figure 1) as it indicates to what extent a node connect s sub-
groups, as a bridge, and then is likely to spread an entity across the
whole network (Newman, 2005).
To date, betweenness has been related to network cohesion
(Lusseau & Newman, 20 04), infection processes (Balasubramaniam
et al., 2016), information transmission (Pasquaretta et al., 2016), sex
(Zhang, Li, Qi, MacIntosh, & Watanabe, 2012), age, rank, kinship
(Bret et al., 2013) and fitness (Gilby et al., 2013). Nodes with the
highest betweenness usually link clusters/modules of nodes within
the networks (e.g. different subgroups or populations) and may thus
have an impor tant role in group cohesion or exchange of entities
(disease, information, genes). However, betweenness is not always
the most informative network measure for an individual's role in
disease spread and such variation could be related to the network
structure (Rodrigues, 2019).
Special attention must be paid regarding the calculation of the
betweenness since the way it is calculated depends on whether
the network is binary or weighted, direc ted or undirected and on
whether the lowest or the highest link/relationship strength is in-
terpreted as the shortest path. Therefore, the different calculations
may lead to dif ferent values. Fur thermore, bet weenness seems to be
very sensitive to sampling effort (Krause et al., 2014).
Closeness is another well-known network measure to study node
centrality but we do not discuss it here as it is ver y similar—although
less frequently used—to betweenness (same variants, same consider-
ations required), and betweenness is usually preferred in ASNA.
1.2.4 | Local clustering coefficient
The local clustering coefficient measures the number of closed tri-
plets* over the total theoretical number of triplets (i.e. open and
closed), where a triplet is an ensemble of three nodes that are con-
nected by either two (open triplet) or three (closed triplet) edges.
This measure aims to examine the links that may exist between the
alters of ego and measures the cohesion of the network (Figure 1).
The main topological effect of closed triplets is the clusterization of
the network, generating cohesive clusters, and is thus strongly re-
lated to modularity (see corresponding section). The local clustering
coefficient can be computed in a binary network by measuring the
proportion of links between the nodes of an ego-network* divided by
the number of potential links between them. In weighted networks,
several versions exist such as those from Barrat, Barthelemy, Pastor-
Satorras, and Vespignani (2004) or Opsahl and Panzarasa (2009). To
date, no attempt has been made in ASNA to evaluate which version
of the clustering coefficient may be the most appropriate accord-
ing to the research question. Therefore, careful attention is needed
when choosing the variant as this may lead to different biological
interpretations. For example, Opsahl's generalized clustering coef-
ficient proposes four variants to consider triplets’ link weights (the
arithmetic or geometric mean or using the weight of the weakest*
or strongest* links). Opsahl's geometric mean variant considers tri-
plet weights heterogeneity (and is robust against extreme values of
weights) whereas Barrat's variant does not. Thus, heterogeneity of
weights should be preferred in social systems with high social het-
erogeneity such as groups with high hierarchy steepness for example.
Finally, the minimum variant (using the weight of the weakest link in
a closed triplet) should be preferred when trying to understand the
mechanisms that shape link creation in animal societies since this
variant helps determine the minimum threshold needed for closed
triplets to appear.
Methods in Ecology and Evoluon
SOSA et Al .
One major asset of this measure is that it is both local and global,
which allows to examine for example how such micro-motifs* affect
the overall network structure (Wharrie, Azizi, & Altmann, 2019). As
we will see, the clustering coefficient examines different aspects of
social networks and animal societies, going from individual hetero-
geneity of social interactions (present section) to the analysis of the
overall group structure (see Global clustering coefficient) and it also
explains patterns in links’ creation (see Transitive triplets). However,
the local and global clustering coefficients can be importantly re-
lated to density so both measures require special attention when
data are collected through GoG and, additionally, density should be
added as factor of control.
1.3 | Examining patterns of node interactions
Patterns of interac tions (how and with whom individuals interact)
can be examined using specific network measures* that analyse
local-scale interactions within a network and make possible to test
hypotheses about the mechanisms underlying network connectiv-
ity (Figure 2). These types of measures are generally used to test
mechanistic biological questions, such as what factors (e.g. eco-
logical as well as sociodemographic) affect individuals’ interactions/
associations. However, because these patterns of interactions are
also known to affect global network features, such as group resil-
ience or reciprocal interactions, and to occur in a wide variet y of ani-
mal taxa, they may be crucial elements within the general processes
that shape animal societies and populations.
1.3.1 | Assortativity
Assor tativity (Newman, 2003) is probably the most used measure
to study homophily (preferential associations or interactions among
individuals sharing the same characteristics; Lazarsfeld & Merton,
1954). Assortativity values range from −1 (total disassortativity
i.e. all the nodes associate or interact with those with the opposite
characteristic, such as males interacting exclusively with females)
to 1 (total assortativity i.e. all the nodes associate or interact with
those with the same characteristic such as males interacting only
with males). The assor tativity coef ficient measures the proportion
of links between and within clusters of nodes with same characteris-
tics. Individuals’ characteristics can be continuous (e.g. age, individ-
ual network measure, personality) or categorical features (e.g. sex,
matriline belonging; Figure 2). Assortativity does not consider direc-
tionalit y* and can be measured in weighted (Leung & Chau, 2007)
or binary (Newman, 2003) networks using categorical or continu-
ous characteristics (Figure 2). The use of one or other assortativit y
variant depends on the type of characteristics being examined and,
whenever possible, the weighted version should be preferred since it
is more reliable than the binary version (Farine, 2014).
Recent studies in human research argue that homophily promotes
cooperation, social learning, and cultural and norm transmission among
strangers (Allen, Weinrich, Hoppitt, & Rendell, 2013). Homophily ac-
cording to different phenotypes such as sex, age, kinship, hierarchical
rank (Sosa, 2016), degree (Croft et al., 2005), personality (Croft et al.,
2009) or body size (Leu et al., 2016) has been found in several species
inc luding fish (Crof t et al., 200 5), birds (Johnso n et al ., 2017), cet acean s
(Hunt, Allen, Bejder, & Parra, 2019), humans (Wang, Suri, & Watts,
2012) and other mammals ( Williamson, Franks, & Curley, 2016). The
fact that similar homophilic mechanisms are found in a wide range of
taxa suggests that homophily may have been a driver for cooperation
between congeners (Apicella, Marlowe, Fowler, & Christakis, 2012).
One question that remains open, however, is whether assortativity is a
consequence of evolution or a prior condition for cooperation, which
would need to be investigated fur ther.
1.3.2 | Transitive triplets
Transitive triplets are micro-motifs that have widely been widely ex-
amined in ASNA in recent years. Transitive triplets are closed triplets
where the links among the nodes follow a specific temporal pattern
of creation, that is when the establishment of links between nodes
A and B and between nodes A and C is followed by the establish-
ment of a link between nodes B and C. This network measure can
be computed in directed, binary or weighted networks. This type
of connections can be studied over time based on the creation of
links. From a static perspec tive, directionality can be considered by
calculating the number of transitive triplets divided by the number
of potential transitive triplets, and weights can also be considered by
using Opsahls’ variants, which are discussed in the section on local
clustering coefficient (Opsahl & Panzarasa, 2009). While transitiv-
ity is importantly related to the clustering coefficient (the clustering
coefficient includes transitive triplets), not all close triplets are tran-
sitive. Transitive triplets are one of the 16 possible configurations of
a triplet considering open and closed triplets as well as link direction-
ality (i.e. triad census).
Transitive triplets have been used in animal affiliative social net-
works (Borgeaud, Sosa, Bshary, Sueur, & Waal, 2016; Boucherie,
Sosa, Pasquaret ta, & Dufour, 2016; Ilany, Booms, & Holekamp, 2015;
Sosa, Zhang, & Cabanes, 2017; Waters & Fewell, 2012) to highlight
‘triadic closure’, commonly described as ‘the friend of my friend is
my friend’. Ilany et al. (2015) evidenced that several factors (rainfall,
prey availability, sex, social rank, dispersal status and topological ef-
fects) shape social dynamics in wild hyenas. Among all these factors,
transitive triplets appeared as the most consistent and social dynam-
ics (link creation) could not be explained without it. This micro-motif
represents an interesting measure when studying social network
resiliency and efficiency. For example, in ants, transitive triplets
appear supporting the hypothesis of adapted and selected pat-
terns of interactions to increase colony functionality and efficiency
(Waters & Fewell, 2012). Moreover, the main topological effect of
triadic closure is the clusterization of the network generating cohe-
sive groups and it seems to be closely linked to the emergence of
reciprocity, altruism and cooperation (Davidsen, Ebel, & Bornholdt,
Methods in Ecology and Evoluon
SOSA et Al .
2002). As for assor tativity, studies testing how this micro-motif
affects the spread of information could help gain knowledge on this
crucial mechanism in the evolution of animal societies.
Transitive triplets have also been used to study agonistic networks
and animals’ dominance hierarchy. For instance, the study of Dey and
Quinn (2014) showed that pukeko agonistic net works emerge from
both individual characteristics and endogenous self-organization of
dominance relationships (i.e. transitive triplets). While triad census has
not been widely used in the past, few studies have started to use these
micro-motifs to examine hierarchy linearity on the basis of occurrence
of reciprocal triplets for example (Shizuka & McDonald, 2012).
Transitive triplets, and triad census more generally, help to
understand how relationships between individuals emerge and
change over time and how these changes may be a consequence of
changes in others’ relationships (Figure 2). The studies mentioned
above investigated triplets’ configuration using unweighted* net-
works. While the weighted variant of transitive triplets (Opsahl &
Panzarasa, 2009) may allow researchers to better understand and
predict how and when links bet ween two individuals are created, it
remains unused in ASNA to date.
1.4 | Examining network structure and properties
The structure of this section is based on the distinction between
network connectivity and social diffusion (information or disease
spread). Both of these aspects may overlap the use of the network
measures that quantify the m (Figure 3). However, the social d iffusion
section contains measures specifically designed to study theoretical
(i.e. considering the diffusion is perfectly related to network links
and link weights) social diffusion features based on the geodesic
distances (see corresponding section). Aspects of the struc ture and
proper ties of a group (e.g. cohesion, sub-grouping) can be quanti-
fied using global network measures*. For instance, one may quantify
proper ties such as network resilience* (see Diameter), network clus-
terization* (see Modularity) through network connectivity analysis,
or network transmission efficiency* (see Global efficienc y) through
network theoretical social diffusion analysis (Figure 3). These differ-
ent network structures have been used in ASNA to study dif ferent
evolutionary as how the network is structured, resilient or efficient
(Puga-Gonzalez, Sosa, & Sueur, 2018) and ecological questions as
how ecological factors such as pathogens affect the network struc-
ture (Croft et al., 2011).
1.4.1 | Examining network connectivity
Network connectivity can be studied using global network meas-
ures that describe the cohesion of the network and how this cohe-
sion may be affected by intrinsic (e.g. species social organization and
structure) or extrinsic factors (e.g. ecological factors as pathogens).
There are three main measures for connectivity discussed in this sec-
tion: density, modularity and clustering coefficient. A s mentioned
above, all these measures may affect social diffusion as high density
and clustering coef ficient induce a fast rate whereas high modularit y
induces a low rate of spread.
The density is the ratio between existing links and all potential links
of a network. This measure is easy to interpret, it assesses how a
network is fully connected. Density does not consider directionality
neither link weights.
In ASNA , a link has been found between density and factors
such as living condition (i.e. higher densit y in captive groups than
in wild groups), group size (i.e. Balasubramaniam et al., 2017; with
the larger the group, the lower the density), seasonality (i.e. higher
density during the mating season; Brent, MacLarnon, Platt, &
Semple, 2013), habitat structural complexity (i.e. higher density
in complex habitats; Leu et al., 2016) and population stress due
to environmental changes (Dufour, Sueur, Whiten, & Buchanan-
Smith, 2011).
However, cautions should be taken when studying density
since this measure may depend on the biology of the species (e.g.
social system and group size) and because several other network
measures appear correlated with it. Density is correlated with de-
gree distribution (see corresponding section), geodesic distances
(see corresponding section) and the frequency of micro-motifs,
like closed triplets* and thus clustering coefficients (see corre-
sponding section; Rankin et al., 2016). These correlations between
density and other global net work measures make it necessar y to
control for network density when comparing global network mea-
sures from different groups, conditions or species. Furthermore,
when comparing species, special attention should be put that the
social organizations (e.g. group size, sex ratio) are equivalent and
thus that interspecies comparisons are meaningful. Furthermore,
the type of behaviour (the rarer the behaviour, the lower the den-
sity; Castles et al., 2014), the size of the network and the sampling
effort are other factors that may influence density and should be
taken into consideration when comparing networks. Methods to
control for such biases have already been proposed (e.g. evalua-
tion of the data collec tion robustness) and should be used when-
ever differences in global network measures (density or other
ones) are assessed (Balasubramaniam et al., 2017). Another option
is to use weighted network measures that are theoretically less
correlated with network density.
Modularity is a measure designed to quantify the degree to which
a network could be divided into different groups or clusters and its
value ranges from 0 to 1. Networks with high modularity have dense
connections within the modules but sparse connections between
the modules. Modularity can be computed in weighted, binar y, di-
rected or undirected networks.
It has been evidenced that modularity varies according to dom-
inance st yle in macaque species, with higher modularity found
in despotic species (Sueur et al. 2011). Fission–fusion societies as
Methods in Ecology and Evoluon
SOSA et Al .
elephants (Wittemyer & Getz, 2007), geladas (Matsuda et al., 2015)
or snub-nosed monkeys (Zhang et al., 2012) show many units and
thus high modularity compared to cohesive groups. Modularity also
seems to be linked to evolutionary advantages such as greater co-
operation by the creation of clusters of cooperators (Marcoux &
Lusseau, 2013) or reduced risks of transmission of pathogens by
decreasing associations between clusters (Nunn, Jordán, McCabe,
Verdolin, & Fewell, 2015). Individuals that interlink the different
clusters may be those with specific social status as observed in dol-
phins (Lusseau & Conradt, 20 09) but clusters can also be linked by
weak links that allow to maintain a certain cohesion and social trans-
mission as described in giraffes (VanderWaal et al., 2016).
Several algorithms have been proposed to identify the different
clusters in a network. These can be categorized according to the
process used to identify the clusters such as spectral optimization
(leading eigenvector), based on the struc ture of the edges (edge be-
tweenness), or modularity optimization (Fastgreedy or Louvain algo-
rithm). For an overview, see Yang, Algesheimer, and Tessone (2016).
Until recently, no research had investigated what would be the im-
pact of choosing different community detection algorithms in the
results (Aldecoa & Marín, 2013; Sumner, McCabe, & Nunn, 2018).
Sumner et al. (2018) showed possible variations between those dif-
ferent algorithms; therefore, we recommend to choose carefully an
appropriate community detection algorithm for the question of in-
terest. Unfortunately, it is only recently that these questions have
been addressed and a general guideline cannot be provided except
that multiple algorithms may be used and the results may be com-
pared. Also note that such precautions could apply to any clusteri-
zation algorithm.
Global clustering coefficient
The global clustering coefficient, like the local clustering coeffi-
cient, evaluates how well the alters of ego are interconnected and
measures the cohesion of the network. It s main topological effect
is the clusterization of the network, generating cohesive clusters,
and is thus strongly related to modularit y. However, it becomes
highly correlated with density and less to modularity as the den-
sity grows. Several variants of the global clustering coefficient can
be found: (a) the ratio of closed triplets to all triplets (open and
closed), (b) the binary local mean clustering coef ficient that de-
rives from the node level (see Local clustering coefficient). The
binary local mean clustering coefficient allows to consider node
heterogeneity and thus should be preferred over the first variant.
Weighted versions also exist and are based on the same variants
described in the section on the local clustering coefficient and re-
quire the same considerations.
1.4.2 | Examining social diffusion
One major aspect that SNA brings in the study of social struc ture is the
possibility to examine social diffusion of disease, information transmis-
sion, new behaviour or ecosystems’ food flow in a network (Figure 3).
One of the measures that make this possible is the geodesic distance
and derived measures such as global efficiency and diameter. While
geodesic distance is not often used in ASNA, it is essential for calculat-
ing other network measures such as diameter, global efficiency, node
betweenness (see corresponding sections). Therefore, we discuss geo-
desic distance in this section to inform the reader that the cautions
needed when computing geodesic distances must also be considered
when calculating its derived network measures.
Geodesic distance
Geodesic distance is the shortest path considering all potential
dyads in a network. This measure thereby evidences the fastest path
of diffusion. Despite its usefulness in the study of epidemiology,
geodesic distance remains seldomly used in ASNA due to its high
sensitivity to observation biases such as unobserved interactions or
misidentification of individuals (Krause et al., 2014). Geodesic dis-
tance can be calculated in binary, weighted*, directed or undirected
networks. In weighted networks, it can be normalized (by dividing all
links by the network weight means) and the strongest or the weakest
links can be considered as the fastest route between two nodes. This
great number of variants of geodesic distance can greatly affect the
results and interpretations. Researchers must thus have knowledge
of the variants and which one is the most appropriate according to
their research question (Opsahl, Agneessens, & Skvoretz, 2010).
For example, many software calculate the geodesic distance
using the paths with the lowest weights as the shortest paths be-
cause they were designed for research related to transportation
routes or information transmission (e.g. road transportation or
internet connection). However, in ASNA, the links with the high-
est weight s are usually those of greater interest as they represent
preferential interactions/associations. For example, the probability
to learn a new behaviour may be higher between individuals that
are more frequently in contact or close proximity (Hoppitt & Laland,
2013). Yet, the weakest links can also be of interest for questions
related to epidemiology. For example, although a pathogen is more
likely to be transmitted among individuals sharing strong links, weak
links may still play a role in disease transmission (VanderWaal et al.,
2016). Directionality is also an important variant to consider when
examining if diffusion can only follow a certain directionality such
as pathogens that can be transmitted only by individuals carrying it.
Global efficiency
Global ef ficiency is the ratio between the number of individuals and
the number of connections multiplied by the network diameter. It
provides a quantitative measure of how efficiently information is ex-
changed within the nodes of the network. As global efficiency gives
a probability of social diffusion, it may help better understand so-
cial transmission phenomena in short-term and long-term (Migliano
et al., 2017). Pasquaretta et al. (2014) found a positive correlation
between the neocortex ratio and the global efficiency in primate
species with a higher neocortex ratio. By drawing a parallel between
cognitive capacities and social network efficiency, this study evi-
denced that in species with higher neocor tex ratio, individuals may
Methods in Ecology and Evoluon
SOSA et Al .
adjust their social relationships in order to gain better access to so-
cial information and thus optimize network efficiency. Alternatively,
studies on epidemiology in ant colonies showed that ants adapt their
interaction rate to decrease the network efficiency when infected
by a pathogen (Stroeymey t et al., 2018).
The diameter of a network represents the longest path of the short-
est paths in the network. Diameter is used in ASNA to examine the
aspects such as network cohesion, the rapidness of information or
disease transmission. While global efficiency measures the theoreti-
cal social diffusion spread, diameter informs on the maximum paths
of diffusion to reach all nodes.
While diameter was first used in the social sciences to study
information diffusion (Milgram, 1967), in ASNA it is mostly used
to examine social cohesion, and the resilience of the net work co-
hesion to the removal of a certain amount of central individuals
(Lusseau, 2003; Manno, 2008; Sosa, 2014; Williams & Lusseau,
2006). However, further investigation may be needed to test if the
removal of central individuals gives a fair picture of biological group
resilience proper ties since currently these analyses do not account
for the creation of new links after the loss of individuals and de-
mographic variations (Firth et al., 2017). If future outcomes support
this deletion simulation assumption, studies based on a comparative
analysis may represent an interesting research approach to under-
stand how natural selection may have favoured resilience properties
in some species while it has not in others. For example, we could
expect variation according to group structure (higher resilience in
stable matriline groups than in fusion–fission societies). Moreover,
given the insight that these simulations could provide into group or
ecosystem resilience properties, those may be of great interest for
conservation purposes (Delmas et al., 2019).
This updated overview of the most commonly used network meas-
ures in ASNA highlights the increasing prominence of techniques
deriving from graph theor y, as well as the insights they brought and
their diversity. Some of these techniques were developed in specific
contexts and for well-defined questions (e.g. Latora & Marchiori,
2001 about global efficiency in neurology). It is very appealing to
reuse them with different focus although this would require a thor-
ough understanding of the mathematical background in order to
know what is being measured and to decide whether a given meas-
ure applies or not to the question raised.
We hope that this non-exhaustive overview will contribute to
facilitate future research in ASNA by helping investigators select
the most relevant network measure and variant according to their
research question. Moreover, we would like to point out that when
using SNA , one is often led to test multiple measures for a single re-
search question as these may reveal different aspects of individuals’
socialit y (direct or indirect links for example). However, it is worth
mentioning that all these measures are computed from the same
mathematical object (the network) and can therefore be correlated
(Bounova & De Weck, 2012). This correlation may be low or high
according to different parameters affecting the network, as the spe-
cies social system or organization, its size, etc. While this has been
discussed punctually along the manuscript, we cannot detail here all
the possible autocorrelations between network measures as this is
case-specific and would fall out of the scope. Nonetheless, we may
recommend to run correlation tests prior to the analyses or to use
the variance inflated factor to control for such bias in correlation
Continuous advances in graph theory such as graph signal pro-
cessing (Shuman, Narang, Frossard, Ortega, & Vandergheynst, 2013)
or multi-layer networks (Kivelä et al., 2014) will undoubtedly give
rise to novel measures with new applications in ASNA . With this per-
spective in mind, investigators need to make constant effor t testing
different versions of measures, clearly stating the mathematical in-
terpretations and what is exac tly being measured, expounding their
streng ths and limits and explaining why chose this variant rather
than another in order for others to apprehend their relevance de-
pending on the context.
We would like to thank Vincent Viblanc and reviewers from Methods
in Ecology and Evolution whose useful comments have substantially
enhanced the qualit y of the manuscript.
S.S. listed all metrics’ variants and wrote the first draft of the manu-
script; I.P.-G. and C.S. participated in the writing of the final version.
This manuscript does not contain any data or code.
Sebastian Sosa
Cédric Sueur
Ivan Puga-Gonzalez
Aldecoa, R., & Marín, I. (2013). Exploring the limits of communit y
detection strategies in complex networks. Scientific Repor ts, 3(1) , 1–11 .
Allen, J., Weinrich, M., Hoppitt, W., & Rendell, L. (2013). Net work-based
diffusion analysis reveals cultural transmission of lobtail feeding in
humpback whales. Science, 340, 485–488.
scien ce.1231976
Almeling, L., Hammerschmidt , K., Sennhenn-Reulen, H., Freund, A. M., &
Fischer, J. (2016). Motivational shifts in aging monkeys and the ori-
gins of social selectivity. Current Biology, 26, 174 4–1749. ht tps ://do i.
Apicella, C. L., Marlowe, F. W., Fowler, J. H., & Christakis, N. A. (2012).
Social networks and cooperation in hunter-gatherers. Nature, 481,
497. htt ps:// /10.1038/nat ur e10736
Aplin, L . M., Farine, D. R., Mann, R. P., & Sheldon, B. C. (2014). Individual-
level personality influences social foraging and collective behaviour
in wild birds. Proceedings of the Royal Society B, 281, 20141016.
Methods in Ecology and Evoluon
SOSA et Al .
Balasubramaniam, K. N., Beisner, B. A., Berman, C. M., De Marco, A.,
Dub oscq, J., Ko irala , S., … Mol es ti, S. (2 017). Th e inf luence of phylog-
eny, social st yle, and sociodemographic fac tors on macaque social
network structure. American Journal of Primatology, 80, e22727.
Balasubramaniam, K., Beisner, B., Vandelees t, J., Atwill, E., & McCowan,
B. (2016). Social buffering and contact transmission: Net work con-
nections have beneficial and detrimental effects on Shigella infec-
tion risk among captive rhesus macaques. P ee rJ, 4, e263 0.
Barrat , A., Barthelemy, M., Pastor-Satorras, R ., & Vespignani, A. (2004).
The architecture of complex weighted networks. Proceedings of the
National Academy of Sciences of the United States of America, 101,
3747–3752. https://doi.or g/10.1073/pnas.0 40 00 87101
Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). Ucinet for Windows:
Software for social network analysis. Harvard, MA: Analytic Technologies.
Borgat ti, S. P., Everett, M. G., & Johnson, J. C. (2018). Analyzing social
networks. Los Angele s, CA: Sage.
Borgeaud, C., Sosa, S., Bshary, R., Sueur, C., & Van de Waal, E. (2016).
Intergroup variation of social relationships in wild vervet monkeys:
A dynamic network approach. Frontiers in Psychology, 7, 915. https://
Borgeaud, C., Sosa, S., Sueur, C., & Bshary, R. (2017). The influence of
demographic variation on social network stabilit y in wild vervet
monkeys. Animal Behaviour, 134, 155–165. htt ps://
j.anbeh av.2017.09.028
Boucherie, P. H., Sosa, S., Pasquaretta, C., & Dufour, V. (2016). A longitudi-
nal network analysis of social dynamics in rooks Corvus frugilegus:
Repeated group modifications do not affect social network in cap-
tive rooks. Current Zoology, 63, 379–388.
Bounova, G., & De Weck, O. (2012). Overview of metrics and their cor-
relation patterns for multiple-metric topology analysis on heteroge-
neous graph ensembles. Physical Review E, 85, 016117. ht tps://doi.
org /10.1103/P hysR evE. 85.016117
Brent, L. J., MacLarnon, A., Plat t, M. L., & Semple, S. (2013). Seasonal
changes in the structure of rhesus macaque social net works.
Behavioral Ecology and Sociobiology, 67, 349–359.
Brent, L ., Ruiz-Lambides, A., & Platt, M. (2017a). Family network size
and survival across the lifespan of female mac aques. Proceedings of
the Royal Society B: Biological Sciences, 284, 20170515. https://doi.
org /10.10 98/rspb .2017.0515
Brent, L., Ruiz-Lambides, A., & Platt, M. (2017b). Persistent social isolation
reflects identity and social context but not maternal effects or early envi-
ronment. Scientific Reports, 7, 17791.
Brent, L ., Semple, S., Dubuc, C ., Heistermann, M., & MacL arnon , A.
(2011). Social capital and physiological stress levels in free-ranging
adult female rhesus macaques. Physiology & Behavior, 102, 76–83. eh.2010.09.022
Bret, C., Sueur, C., Ngoubangoye, B., Verrier, D., Deneubourg, J.-L., &
Petit, O. (2013). Social structure of a semi-free ranging group of
mandrills (Mandrillus sphinx): A social network analysis. PLoS ONE, 8,
e83015. al.pone.0083015
Castles, M., Heinsohn, R., Mar shall, H . H., Lee, A. E., Cowlishaw, G., &
Carter, A. J. (2014). Social networks created with different tech-
niques are not comparable. Animal Behaviour, 96, 59–67. https://doi.
org/10.1016/j.anbeh av.2014.07.023
Croft, D. P., Edenbrow, M., Darden, S. K., Ramnarine, I. W., van
Oosterhout, C., & Cable, J. (2011). Effect of g yrodactylid ectopar-
asites on host behaviour and social net work structure in guppies
Poecilia reticulata. Behavioral Ecology and Sociobiology, 65, 2219–2227.
Croft, D., James, R., Ward, A., Botham, M., Mawdsley, D., & Krause, J.
(2005). Assortative interactions and social networks in fish. Oecologia,
143, 211–219.
Croft, D. P., Krause, J., Darden, S. K., Ramnarine, I. W., Faria, J. J., & James,
R. (2009). Behavioural trait assortment in a social network: Patterns
and implications. Behavioral Ecology and Sociobiology, 63, 1495–1503.
Csardi, G., & Nepusz, T. (2006). The igraph soft ware package for com plex
network research. InterJournal, Complex Systems, 1695, 38.
Davidsen , J., Ebel, H., & Bornholdt, S. (20 02). Emergence of a small
world from local interactions: Modeling acquaintance networks.
Physical Review Letters, 88, 128701.
R evLett.88.128701
Delmas , E., Besson, M. , Brice, M.-H., Bur kle, L. A., Da lla Riva, G. V., Fort in,
M.-J., … Poisot, T. (2019). Analysing ecological networks of species
interactions. Biological Reviews, 94, 16–36. https://doi.o rg /10.1111 /
br v.124 33
Dey, C. J., & Quinn, J. S. (2014). Individual at tributes and self-
organizational processes affect dominance network structure in
pukeko. Behavioral Ecology, 25(6), 1402–1408.
Dufour, V., Sueur, C., Whiten, A ., & Buchanan-Smith, H. M. (2011). The
impact of moving to a novel environment on social networks, activ-
ity and wellbeing in two new world primates. American Jou rnal of
Primatology, 73, 802–811.
Farine, D. R. (2014). Measuring phenotypic assortment in animal social
networks: Weighted associations are more robust than binary edges.
Animal Behaviour, 89, 141–153. av.
Firth, J. A., Voelkl, B., Crates, R. A., Aplin, L . M., Biro, D., Croft, D. P., &
Sheldon, B. C. (2017). Wild birds respond to flockmate loss by in-
creasing their social network associations to others. Proceedings of
the Royal Society B: Biological Sciences, 284, 20170299. https://doi.
org /10.10 98/rspb .2017.029 9
Franks, D. W., Ruxton, G. D., & James, R. (2010). Sampling animal association
networks with the gambit of the group. Behavioral Ecology and Sociobiol-
ogy, 64, 493–503.
Gilby, I. C., Brent, L. J., Wroblewski, E. E., Rudicell, R. S., Hahn, B. H.,
Goodall, J., & Pusey, A. E. (2013). Fitness benefit s of coalitionary ag-
gression in male chimpanzees. Behavioral Ecology and Sociobiology, 67,
Hoppitt, W., & Laland, K. N. (2013). Social learning: An introduction to mech-
anisms, methods, and models. Princeton, NJ: Princeton University Press.
Hunt, T. N., Allen, S. J., Bejder, L., & Parra, G. J. (2019). Assor tative in-
teractions revealed in a fission–fusion society of Australian hump-
back dolphins. Behavioral Ecology, 30(4), 914–927. https://doi.
org/10.1093/behec o/arz029
Ilany, A., Booms, A. S., & Holekamp, K. E. (2015). Topological effects of net-
work structure on long-term social network dynamics in a wild mam-
mal. Ecology Letters, 30, 914–927.
Johnson, K. V.-A., Aplin, L. M., Cole, E. F., Farine, D. R., Firth, J. A., Patrick,
S. C., & Sheldon, B. C. (2017). Male great tits assort by personality
during the breeding season. Animal Behaviour, 128, 21–32. https:// av.2017.04.001
Kivelä, M., Arenas, A., Barthelemy, M., Gleeson, J. P., Moreno, Y., & Porter,
M. A. (2014). Multilayer networks. Journal of Complex Networks, 2,
203–271. t/cnu016
Krause, J., James, R ., Franks, D. W., & Croft, D. P. (2014). Animal social
networks. Oxford, UK: Oxford University Press.
Latora , V., & Marchiori, M. (2001). Efficient behavior of small-world net-
works. Physical Review Letters, 87, 198701. http s://
PhysR evLett.87.198701
Lazar sfeld, P. F., & Merton, R. K. (1954). Friendship as a social proce ss:
A substantive and methodological analysis. Freedom and Control in
Modern Society, 18, 18–66.
Leu, S. T., Farin e, D. R., Wey, T. W., Sih, A., & Bull , C. M. (2016). Enviro nment
modulates population social structure: Experimental evidence from
replicated social networks of wild lizards. Animal Behaviour, 111,
23–31. av.2015.10.001
Methods in Ecology and Evoluon
SOSA et Al .
Leung, C ., & Chau, H. (2007 ). Weighted assortative and disassor tative
networks model. Physica A: Statistical Mechanics and its Applications,
378, 591–60 2. http s://doi.o rg/10.1016/j .physa.20 06.12.0 22
Liao, Z., Sosa, S., Wu, C., & Zhang, P. (2018). The influence of age on wild
rhesus macaques' affiliative social interactions. American Journal of
Primatology, 80, e22733.
Lusseau, D. (2003). The emergent properties of a dolphin social network.
Proceedi ngs of the Royal Society of London. Series B: Biological Sciences,
270, S186–S188.
Lusseau, D., & Conradt, L. (2009). The emergence of unshared consensus
decisions in bottlenose dolphins. Behavioral Ecology and Sociobiology,
63, 1067–1077. https://doi.or g/10.10 07/s00265-0 09-0740-7
Lusseau, D., & Newman, M. E. (2004). Identifying the role that animals play in
their social networks. Proceedings of the Royal Society of London B: Biolog-
ical Sciences, 271, S477–S481.
Manno, T. G. (20 08). Social net working in the Columbian ground squirrel,
Spermophilus columbianus. Animal Behaviour, 75, 12211228. https:// av.2007.09.025
Marcoux , M., & Lusseau, D. (2013). Network modularity promotes co-
operation. Journal of Theoretical Biology, 324, 103–108. h ttps://doi.
Matsuda, I., Fukaya, K., Pasquaretta, C ., & Sueur, C. (2015). Factors influ-
encing grooming social networks: insights from comparisons of colo-
bines with different dispersal patterns. In Dispersing primate females
(pp. 231–254). Tokyo, Japan: Springer Verlag.
Migliano, A. B., Page, A. E., Gómez-Gardeñes, J., Salali, G. D., Viguier, S.,
Dyble, M., … Vinicius, L. (2017). Characterization of hunter-gatherer
networks and implications for cumulative culture. Nature Human
Behaviour, 1, 0043.
Milgram, S. (1967). The small world problem. Psychology Today, 2, 60–6 7.
Moyers, S. C., Adelman, J. S., Farine, D. R., Moore, I. T., & Hawley, D.
M. (2018). Exploratory behavior is linked to stress physiology and
social network centrality in free-living house finches (Haemorhous
mexicanus). Hormones and Behavior, 102, 105–113. https://doi.
Newman, M. E. (2003). Mixing patterns in net works. Physical Review E,
67, 026126. evE.67.026126
Newman, M. E. (20 05). A measure of betweenness centrality based on
random walks. Social Networks, 27, 39–54.
j.socnet.200 4.11.009
Nunn, C. L., Jordán, F., McCabe, C. M., Verdolin, J. L., & Fewell, J. H. (2015).
Infectious disease and group size: More than just a numbers game.
Philosophical Transactions of the Royal Society B: Biological Sciences,
370, 20 140111.
Opsahl, T., Agneessens, F., & Skvoretz, J. (2010). Node centrality in weighted
networks: Generalizing degree and shortest paths. Social Networks, 32,
Opsahl, T., & Panzarasa, P. (2009). Clustering in weighted network s. Social
Networks, 31, 155–163.
Pasquaretta, C., Bat testi, M., Klenschi, E., Bousquet, C. A ., Sueur, C.,
& Mery, F. (2016). How social network struc ture affects decision-
making in Drosophila melanogaster. Proceedings of the Royal Society. B,
Biological Sciences, 283, 20152954.
Pasquaretta, C., Levé, M., Claidière, N., van de Waal, E ., Whiten, A.,
MacIntosh, A . J. J., … Sueur, C. (2014). Social networks in primates:
Smart and tolerant species have more efficient networks. Scientific
Reports, 4, 7600. 7600
Puga-G onzale z, I., Sosa, S., & Sueur, C. (2018). Social style and re silience
of macaques’ networks, a theoretical investigation. Pri mates, 60,
Rankin, R. W., Mann, J., Singh, L., Patterson, E. M., Krzyszczyk, E., &
Bejder, L. (2016). The role of weighted and topological net work in-
formation to under stand animal social net works: A null model ap-
proach. Animal Behaviour, 113, 215–228. ht tps://
j.anbeh av.2015.12.015
Rodrigues, F. A. (2019). Network centr ality: An introduction. In A mathe-
matical modeling approach from nonlinear dynamics to complex systems
(pp. 177–196). Cham, Switzerland: Springer.
Scharf, I., Modlmeier, A. P., Beros, S., & Foitzik, S. (2012). Ant societies
buffer individual-level effects of parasite infec tions. The American
Naturalist, 180, 671–683.
Schülke, O., Bhagavatula, J., Vigilant, L ., & Ostner, J. (2010). Social bonds
enhance reproductive success in male macaques. Current Biology, 20,
Shimada, M., & Sueur, C. (2014). The importance of social play network
for infant or juvenile wild chimpanzees at Mahale Mountains National
Park, Tanzania. American Journal of Primatology, 76, 1025–1036.
Shizuka , D., & McDonald, D. B. (2012). A social network perspective
on measurements of dominance hierarchies. Animal Behaviour, 83,
925–934. av.2012.01.011
Shuman, D. I., Narang, S. K., Frossard, P., Ortega, A ., & Vandergheynst, P.
(2013). The emerging field of signal processing on graphs: Ex tending
high-dimensional data analysis to networks and other irregular
domains. IEEE Sign al Processing Magazine, 30, 83–98. https://doi.
org /10.1109/MS P.2012 .2235192
Silk, J. B., Beehner, J. C., Bergman, T. J., Crockford, C., Engh, A. L.,
Moscovice, L. R., … Cheney, D. L. (2010). Strong and consistent social
bonds enhance the longevity of female baboons. Current Biology, 20,
Sosa, S. (2014). Structural architecture of the social net work of a non-
human primate (Macaca sylvanus): A study of its topology in La Forêt
des singes, Rocamadour. Folia Primatologica, 85, 154–163. https:// /10.1159/00036 0986
Sosa, S. (2016). The influence of gender, age, matriline and hierarchical
rank on individual social position, role and interactional patterns in
Macaca sylvanus at ‘La Forêt des singes’: A multilevel social net-
work approach. Frontiers in Psychology, 7.
Sosa, S., Puga-Gonzalez, I., Hu Feng, H., Zhang, P., Xiaohua, X., & Sueur,
C. (2018). A multilevel statistical toolkit to study animal social net-
works: Animal Net work Toolkit (ANT) R package. bioRxiv. h t t p s ://
doi .or g/10 .1101/3 470 05
Sosa, S., Zhang, P., & Cabanes, G. (2017). Social networks dynamics
revealed by temporal analysis: An example in a non-human pri-
mate (Macaca sylvanus) in “La Forêt des Singes”. American Journal of
Primatology, 79, e22662.
Stanton, M. A., & Mann, J. (2012). Early social network s predict sur-
vival in wild bottlenose dolphins. PLoS ONE, 7, e47508. https://doi.
org/10.1371/journ al.pone.0047508
Stroeymeyt, N., Grasse, A. V., Crespi, A., Mersch, D. P., Cremer, S., &
Keller, L. (2018). S ocial network plasticity decreases disease trans-
mission in a eusocial insect. Science, 362, 941–945. https://doi.
org/10.1126/scien ce.aat4793
Sueur, C., Petit, O., De Marco, A., Jacobs, A., Watanabe, K., & Thierr y, B.
(2011). A comparative net work analysis of social style in macaques.
Animal Behaviour, 82, 845–852. ht tps:// av.
Sueur, C., Romano, V., Sosa, S., & Puga-Gonzalez, I. (2019). Mechanisms
of network evolution: A focus on socioecological factors, interme-
diary mechanisms, and selection pressures. Primates, 60, 167–181.
htt ps:// /10.10 07/s10329-018-0682-7
Sumner, K. M ., McCabe, C. M., & Nunn, C . L. (2018). Network size,
structure, and pathogen transmission: A simulation study compar-
ing different community detection algorithms. Behaviour, 155( 7–9),
VanderWaal, K. L., Obanda, V., Omondi, G . P., McCowan, B., Wang, H.,
Fushing, H ., & Isbell, L. A. (2016). The “strength of weak ties” and
helminth parasitism in giraffe social networks. Behavioral Ecology,
arw035. o/arw035
Methods in Ecology and Evoluon
SOSA et Al .
Wang, J., Suri, S., & Watts, D. J. (2012). Cooperation and assortativity
with dynamic par tner updating. Proceedings of the National Academy
of Science s of the United States of Ame rica, 109, 14363–1436 8. ht tps:// /10.1073/p nas.11208 67109
Waters, J. S., & Fewell, J. H. (2012). Information processing in social insect
networks. PLoS ONE, 7, e40337. al.
Wharrie, S., Azizi, L., & Altmann, E. G. (2019). Micro, Meso, Macro: The
effect of triangles on communities in networks. arXiv preprint arXiv:
Whitehead, H. (20 09). SOCPROG programs: Analysing animal social
structures. Behavioral Ecology and Sociobiology, 63, 765–778. https://
Williams, R., & Lusseau, D. (2006). A killer whale social network is v ul-
nerable to targeted removals. Biology Letters, 2, 497–500. https://doi.
Williamson, C. M., Franks, B., & Curley, J. P. (2016). Mouse so-
cial network dynamics and community structure are associ-
ated with plasticity-related brain gene expression. Frontiers in
Behavioral Neuroscience, 10, 152.
Wittemyer, G., & Getz, W. M. (2007). Hierarchical dominance struc ture
and social organization in African elephants, Loxodonta africana.
Animal Behaviour, 73, 671–681. av.
Yang, Z., Algesheimer, R., & Tessone, C. J. (2016). A comparative analysis
of community detection algorithms on artificial networks. Scientific
Reports, 6, 30750. 0750
Zhang, P., Li, B.-G., Qi, X.-G ., MacIntosh, A. J., & Watanabe, K. (2012).
A proximit y-based social network of a group of Sichuan snub-nosed
monkeys (Rhinopithecus roxellana). International Journal of Primatology,
33, 1081–1095. ht tps:// .1007/s10764- 012-96 08-1
Additional suppor ting information may be found online in the
Supporting Information section.
How to cite this article: Sosa S, Sueur C, Puga-Gonzalez I.
Network measures in animal social network analysis: Their
strengths, limits, interpretations and uses. Methods Ecol Evol.
2020;00:1–12. htt ps:// .1111/2 041-210X.1336 6
... Different mechanisms linking variation in sociality could theoretically be fueled by different aspects of sociality . For example, if fitness effects result from increased access to information, occupying a central/brokering position that connects otherwise unconnected parts of the social structure should be most beneficial (Brent 2015;Sueur et al. 2011a), whereas in societies where fitness effects are driven by coalitionary support in competition for social status, individuals with a few very strong social bonds might have the most beneficial position Sosa et al. 2021) The value of assessing covariation between different sociality metrics extends beyond their application to the questions mentioned above. Other research areas that require measuring individuals' position in affinitive and affiliative social structure include social cognition (Almeling et al. 2016;Platt et al. 2016), comparative psychology (Massen and Koski 2014), social neuroscience (Cacioppo and Cacioppo 2012), and social psychopathology (Bauman and Schumann 2018). ...
... In the following, we will refer to these choices and corresponding behavioral measures as different sociality indices (Box 1). The choice of index may depend on the research question, average frequencies of behaviors, observational data collection methods, and visibility of group members as they affect the accuracy of the behavioral measure (Sosa et al. 2021). ...
... The third concept considered here is indirect connectedness (structural connectedness in Ellis et al. (2019)), i.e., inter-individual variation in social embeddedness resulting from variation in how also an individual's direct partners are connected to other group members (in a "friend-of-a-friend" manner, Brent 2015) and how an individual is positioned relative to all other nodes in the network, i.e., in the social structure resulting from affinitive and affiliative behavior of all group members. Different aspects of indirect connectedness can be construed that relate to different biological phenomena (Sosa et al. 2021), e.g., the friends of friends may provide an extended network of cooperators, whereas a broader perspective is needed to assess individual roles in network diffusion of information or parasites. The link between social bonding, social integration, and indirect connectedness remains poorly resolved and may vary to some extent among species and social groups. ...
Full-text available
It has long been recognized that the patterning of social interactions within a group can give rise to a social structure that holds very different places for different individuals. Such within-group variation in sociality correlates with fitness proxies in fish, birds, and mammals. Broader integration of this research has been hampered by the lack of agreement on how to integrate information from a plethora of dyadic interactions into individual-level metrics. As a step towards standardization, we collected comparative data on affinitive and affiliative interactions from multiple groups each of five species of primates to assess whether the same aspects of sociality are measured by different metrics and indices. We calculated 16 different sociality metrics used in previous research and thought to represent three different sociality concepts. We assessed covariation of metrics within groups and then summarized covariation patterns across all 15 study groups, which varied in size from 5 to 41 adults. With some methodological and conceptual caveats, we found that the number of weak ties individuals formed within their groups represented a dimension of sociality that was largely independent from the overall number of ties as well as from the number and strength of the strong ties they formed. Metrics quantifying indirect connectedness exhibited strong covariation with strong tie metrics and thus failed to capture a third aspect of sociality. Future research linking affiliation and affinity to fitness or other individual level outcomes should quantify inter-individual variation in three aspects: the overall number of ties, the number of weak ties, and the number or strength of strong ties individuals form, after taking into account effects of social network density. Significance statement In recent years, long-term studies of individually known animals have revealed strong correlations between individual social bonds and social integration, on the one hand, and reproductive success and survival on the other hand, suggesting strong natural selection on affiliative and affinitive behavior within groups. It proved difficult to generalize from these studies because they all measured sociality in slightly different ways. Analyzing covariation between 16 previously used metrics identified only three rather independent dimensions of variation. Thus, different studies have tapped into the same biological phenomenon. How individuals are weakly connected within their group needs further attention.
... friend of our friend) also seem to be important for social capital Fowler 2007, 2008). These indirect connections affecting network indices as betweenness (Sosa et al. 2020) may strongly affect the evolution of species Figure 2: Evolution of biological age according to chronological age and network age. The dotted line represents variations observed in reversing caste and solicitations in eusocial insects but may result from intervention on social parameters in humans and other animals. ...
... We then calculated different individual and global measures using the ANTs (Sosa et al. 2018b) and igraph (Csardi and Nepusz 2006) R packages. We avoided measures that were not interpretable with the directionality of the edges (see Mersch, 2016;Sosa et al., 2020;Sueur et al., 2011); instead, we selected those that allowed us to make expectations based on our three aforementioned hypotheses. Table S1 shows the different indices and their associated expectations. ...
... The individual network measures included the degree (the number of edges of a node; i.e. the number of individuals giving or receiving trophallaxes from an ant), strength (the total time of trophallaxes of an ant; here we considered in-strength, the total time of trophallaxes received, and out-strength, the total time of trophallaxes given), betweenness (the number of shortest paths passing by a node; i.e. how many individuals an ant connects) and the clustering coefficient (whether individuals with which an ant exchanges food also exchange food). Detailed explanations of these different measures have been provided by previous reviews and books on animal networks (Whitehead 1997;Croft et al. 2008;Sueur et al. 2011;Sosa et al. 2020) Statistical Analysis ...
Full-text available
My thesis analyses how the social environment influences longevity in various animal species (vertebrates and invertebrates) in a gradient of social complexity (gregariousness, cooperative breeding, eusociality). Thanks to complementary measurements (behavioural observations, oxidative stress, telomere length, proteomics, metabolomics), my work reflects contrasting ageing mechanisms depending on the species studied. For example, in birds, longer telomeres are associated with a favourable social environment and greater longevity. On the contrary, among the 10 species of ants studied, the longest-lived ants had shorter telomeres. Furthermore, the complete and combined study of the proteome and metabolome of ants has highlighted certain mechanisms (e.g., sirtuins, mTOR, anti-cancer mechanisms), the universality of which opens the way to a better understanding of the harmful effect of social stress on human health.
... Strength centrality is calculated as the sum of the weights of the edges in a weighted network . This measure represents the sociality of an individual as it estimates the frequency of its interactions (Sosa et al., 2021a) and it was also used to compute gregariousness at the group level (see Global metrics section). Eigenvector centrality is defined as the first non-negative eigenvector value obtained by transforming an adjacency matrix linearly . ...
... Global metrics: as measures to describe the global structure of the network we calculated network density, gregariousness and Typical Group Size (TGS). Network density calculates the ratio between existing links and all potential links of a network, and it assesses the connection of the network as a whole (Sosa et al., 2021a). Gregariousness represents the tendency of individuals to associate with few or many individuals (Godde et al., 2013). ...
Full-text available
Despite its recognized importance for understanding the evolution of animal sociality as well as for conservation, long term analysis of social networks of animal populations is still relatively uncommon. We investigated social network dynamics in males of a gregarious mountain ungulate (Alpine ibex, Capra ibex) over ten years focusing on groups, sub‐groups and individuals, exploring the dynamics of sociality over different scales. Despite the social structure changing between seasons, the Alpine ibex population was highly cohesive: fission–fusion dynamics lead almost every male in the population to associate with each other male at least once. Nevertheless, we found that male Alpine ibex showed preferential associations that were maintained across seasons and years. Age seemed to be the most important factor driving preferential associations while other characteristics, such as social status, appeared less crucial. We also found that centrality measures were influenced by age and were also related to individual physical condition. The multi‐scale and long‐term frame of our study helped us show that ecological constrains, such as resource availability, may play a role in shaping associations in a gregarious species, but they cannot solely explain sociality and preferential association that are likely also to be driven by life‐history linked physiological and social needs. Our results highlight the importance of long‐term studies based on individually recognizable subjects to help us build on our understanding of the evolution of animal sociality.
... Several metrics have been developed to quantify the position of individuals in social networks, in terms of network centrality (Brent 2015;Ellis et al. 2019;Sosa et al. 2021). We used only one metric of centrality to avoid redundancy and "metric hacking" (Webber et al. 2020). ...
... We used only one metric of centrality to avoid redundancy and "metric hacking" (Webber et al. 2020). We chose to use eigenvector centrality (Bonacich 1987) because it is one of the most commonly employed and robust measures of centrality (Brent 2015;Sosa et al. 2021). In our system, weighted eigenvector centrality was highly related to simpler social network metrics, such as the weighted degree (i.e., sum of all strengths of associations of an individual; Spearman rank correlations between eigenvector centrality and weighted degree > 0.99 for all the time periods). ...
In gregarious animals, social network positions of individuals may influence their life-history and fitness. Although association patterns and the position of individuals in social networks can be shaped by phenotypic differences and by past interactions, few studies have quantified their relative importance. We evaluated how phenotypic differences and familiarity influence social preferences and the position of individuals within the social network. We monitored wild-caught common waxbills (Estrilda astrild) with radio-frequency identifiers in a large mesocosm during the non-breeding and breeding seasons of two consecutive years. We found that social networks were similar, and that the centrality of individuals was repeatable, across seasons and years, indicating a stable social phenotype. Nonetheless, there were seasonal changes in social structure: waxbills associated more strongly with opposite-sex individuals in breeding seasons, whereas in non-breeding seasons they instead assorted according to similarities in social dominance. We also observed stronger assortment between birds that were introduced to the mesocosm at the same time, indicating long-lasting bonds among familiar individuals. Waxbills that had been introduced to the mesocosm more recently occupied more central network positions, especially during breeding seasons, perhaps indicating that these birds had less socially-differentiated associations with flock members. Finally, individual differences in color ornamentation and behavioral assays of personality, inhibitory control, and stress were not related to network centrality or association patterns. Together, these results suggest that, in gregarious species like the common waxbill, social networks may be more strongly shaped by long-lasting associations with familiar individuals than by phenotypic differences among group members.
... This disturbance ecology paradigm moving forward would benefit from the use of wildlife-tracking technology (e.g., GPS collars, accelerometers) and paired with other devices (e.g., camera traps and acoustic recorders) to estimate wildlife state variables in response to disturbance [63]. Studies that integrate disparate data sources could help parameterise social network analyses to inform on inter/intra-species interactions after disturbances [64] and move to more mechanistic explanations of the results found during observational studies. Finally, simulation studies should be used to test the efficacy of management and policy and for informing future field data collection [65,66]. ...
Full-text available
Disturbance ecology refers to the study of discrete processes that disrupt the structure or dynamics of an ecosystem. Such processes can, therefore, affect wildlife species ecology, including those that are important pathogen hosts. We report on an observational before-and-after study on the association between forest clearfelling and bovine tuberculosis (bTB) herd risk in cattle herds, an episystem where badgers (Meles meles) are the primary wildlife spillover host. The study design compared herd bTB breakdown risk for a period of 1 year prior to and after exposure to clearfelling across Ireland at sites cut in 2015–2017. The percent of herds positive rose from 3.47% prior to clearfelling to 4.08% after exposure. After controlling for confounders (e.g., herd size, herd type), we found that cattle herds significantly increased their odds of experiencing a bTB breakdown by 1.2-times (95%CIs: 1.07–1.36) up to 1 year after a clearfell risk period. Disturbance ecology of wildlife reservoirs is an understudied area with regards to shared endemic pathogens. Epidemiological observational studies are the first step in building an evidence base to assess the impact of such disturbance events; however, such studies are limited in inferring the mechanism for any changes in risk observed. The current cohort study suggested an association between clearfelling and bTB risk, which we speculate could relate to wildlife disturbance affecting pathogen spillback to cattle, though the study design precludes causal inference. Further studies are required. However, ultimately, integration of epidemiology with wildlife ecology will be important for understanding the underlying mechanisms involved, and to derive suitable effective management proposals, if required.
... We analyzed both the directed and undirected matrices as weighted networks to maintain the strength of the interactions, given our small per pen group size. We examined direct and indirect social connections between the calves' and their pen mates with three centrality measures at the individual or node level: degree, strength and eigenvector (reviewed in 27 ). Nodal metrics describe each node's position in the network relative to the other nodes 28 . ...
Full-text available
Changes in network position and behavioral interactions have been linked with infectious disease in social animals. Here, we investigate the effects of an experimental disease challenge on social network centrality of group-housed Holstein bull dairy calves. Within group-housed pens (6/group) calves were randomly assigned to either a previously developed challenge model, involving inoculation with Mannheimia haemolytia ( n = 12 calves; 3 calves/group) or a control involving only saline ( n = 12 calves; 3 calves/group). Continuous behavioral data were recorded from video on pre-treatment baseline day and for 24 h following inoculation to describe social lying frequency and duration and all active social contact between calves. Mixed-model analysis revealed that changes in network position were related to the challenge. Compared to controls, challenged calves had reduced centrality and connectedness, baseline to challenge day. On challenge day, challenged calves were less central in the directed social contact networks (lower degree, strength and eigenvector centrality), and initiated contact (higher out-degree) with more penmates, compared to healthy calves. This finding suggests that giving rather than receiving affiliative social contact may be more beneficial for challenged calves. This is the first study demonstrating that changes in social network position coincide with an experimental challenge of a respiratory pathogen in calves.
... Individuals are likely to use homophily as a heuristic when exploring the environment in which learning from others (with similar characteristics) can lead to inefficient exploration (Calacci, 2018). This can be reasoned to that homophily promotes cooperation, social learning, and cultural bindings (Sosa et al., 2020). Despite these evidences, little is known about how avatar homophily in terms of attitude and background can influence individuals' exploratory behaviour in a virtual learning environment. ...
Full-text available
Virtual learning environments have been recognized as an area of particular importance by which educators can use to improve desirable learning behaviours. Investigating the impact of different virtual environments on learners’ behaviours has become the centre of attention of researchers, especially during COVID-19. The homophily effect of avatar-identity on individuals’ perceptions of an environment can be a key for understanding their learning behaviours. This study examined the relationship between key constructs related to avatar homophily (background and attitude) and learners’ flow and exploratory behaviour. An online survey was distributed to 157 students (93 males and 64 females with age ranging from 19 to 21 years) who took part in an online learning activity using an avatar-mediated environment (Second Life). The results showed that users’ flow experience can be influenced by the function of perceived background and attitude homophily in an avatar-mediated environment. Flow experience was found to mediate the relationship between avatar homophily and learners’ exploratory behaviour. This study offers a conceptual understanding of the relationship between homophily and individual’s flow state.
... whether a dolphin's associates were themselves closely associated). We chose these two metrics because they better represent dolphin individual sociality and are considered important for survival and reproduction while being less influenced by the individuals' age and the network density (Sosa et al., 2021). ...
Food provisioning promotes close interaction with wildlife but can negatively impact the targeted species. Repeated behavioural disruptions have the potential to negatively impact vital rates and have population level consequences. In Bunbury, Western Australia, food-provisioned female bottlenose dolphins, Tursiops aduncus, suffer reduced reproductive success via lower calf survival. However, the proximal causes of this long-term negative effect remain unknown. To infer processes that could lead to fitness costs, we combined network analyses, Markov Chain, regression models and kernel density estimates to evaluate the social environment, behavioural budget and home range size of provisioned dolphins relative to their nonprovisioned counterparts. We found that provisioned dolphins spent significantly less time socializing and had smaller home ranges and weaker social associations than the nonprovisioned dolphins. Overall, these findings suggest that provisioned dolphins experience a more restricted social environment among themselves, which likely results from investing time in an unnatural foraging tactic around the provisioning site, in proximity to human activities. This modified social environment associated with food provisioning and begging behaviour, reinforced by the limited time spent socializing, could affect the opportunities of calves of provisioned females to acquire fitness-enhancing skills and form essential social bonds. This study highlights the need to consider the potential impact of human activities on the social environment of animals.
... We used the eigenvector centrality as a measure of social centrality. This measures the number and strength of associations of an individual, whilst taking into account the strength of associations of the individuals with whom it is itself associated and is often described as popularity [52]. ...
Full-text available
Stress is the body’s response to cope with the environment and generally better survive unless too much chronic stress persists. While some studies suggest that it would be more stressful to be the dominant individual of the group, others support the opposite hypothesis. Several variables can actually affect this relationship, or even cancel it. This study therefore aims to make the link between social status and the basal level of stress of 14 wild European bison (Bison bonasus, L. 1758) living together. We collected faeces and measured the faecal glucocorticoid metabolites (FGM). We showed that FGM is linked to different variables of social status of European bison, specifically age, dominance rank, eigenvector centrality but also to interactions between the variables. Preferential leaders in bison, i.e., the older and more dominant individuals which are more central ones, are less stressed compared to other group members. Measurement of such variables could thus be a valuable tool to follow and improve the conservation of species by collecting data on FGM and other social variables and adapt group composition or environmental conditions (e.g., supplement in food) according to the FGM concentration of herd individuals.
Full-text available
Ce travail s’inscrit dans l’étude des origines évolutives du langage, par la recherche de propriétés langagières dans la communication gestuelle et multimodale de primates cercopithécidés en captivité, les mangabeys à collier. Par une double approche observationnelle et expérimentale, nous avons montré que les gestes des mangabeys remplissent les critères de définition d’une communication intentionnelle, et peuvent être produits de manière flexible dans différents contextes. Nos observations fournissent également de premiers éléments en faveur d’une intentionnalité des expressions faciales des cercopithécidés, souvent considérées comme de simples indices d’état émotionnel. Cette propriété sociocognitive langagière pourrait ainsi être plus ancienne que ce que nous pensions dans l’histoire évolutive des primates, et être héritée de la communication gestuelle des ancêtres des catarrhiniens, il y a environ 29 millions d’années. De plus, nous avons mis en évidence un effet significatif du contexte interactionnel sur la latéralité gestuelle des mangabeys, suggérant une importance particulière de facteurs sociaux dans l’émergence d’une spécialisation hémisphérique pour la communication intentionnelle, dont le langage humain. Enfin, par une méthode originale, reposant sur des analyses de séquences et de réseau, nous avons décrit la communication multimodale et multicomposante des mangabeys à collier, et montré qu’ils combinent de manière flexible différents types et modalités de signaux en fonction du contexte et de facteurs sociodémographiques. Nos résultats soulignent l’importance d’une approche multimodale pour comprendre la complexité de la communication des primates, et apportent de premier éléments de compréhension sur la fonction des combinaisons de signaux. De futures comparaisons à d’autres espèces et dans différents environnements pourraient permettre d’affiner nos connaissances quant aux possibles contraintes évolutives ayant favorisé une telle complexité de la communication des primates humains et non-humains.
Full-text available
Meso-scale structures (communities) are used to understand the macro-scale properties of complex networks, such as their functionality and formation mechanisms. Micro-scale structures are known to exist in most complex networks (e.g., large number of triangles or motifs), but they are absent in the simple random-graph models considered (e.g., as null models) in community-detection algorithms. In this paper we investigate the effect of micro-structures on the appearance of communities in networks. We find that alone the presence of triangles leads to the appearance of communities even in methods designed to avoid the detection of communities in random networks. This shows that communities can emerge spontaneously from simple processes of motiff generation happening at a micro-level. Our results are based on four widely used community-detection approaches (stochastic block model, spectral method, modularity maximization, and the Infomap algorithm) and three different generative network models (triadic closure, generalized configuration model, and random graphs with triangles).
Full-text available
Understanding individual interactions within a community or population provides valuable insight into its social system, ecology, and, ultimately, resilience against external stimuli. Here, we used photo-identification data, generalized affiliation indices, and social network analyses to investigate dyadic relationships, assortative interactions, and social clustering in the Australian humpback dolphin (Sousa sahulensis). Boat-based surveys were conducted between May 2013 and October 2015 around the North West Cape, Western Australia. Our results indicated a fission-fusion society, characterized by nonrandom dyadic relationships. Assortative interactions were identified both within and between sexes and were higher among members of the same sex, indicating same-sex preferred affiliations and sexual segregation. Assortative interactions by geographic locations were also identified, but with no evidence of distinct social communities or clusters or affiliations based on residency patterns. We noted high residency among females. Models of temporal patterns of association demonstrated variable levels of stability, including stable (preferred companionships) and fluid (casual acquaintances) associations. We also demonstrated some social avoidance. Our results point to greater social complexity than previously recognized for humpback dolphins and, along with knowledge of population size and habitat use, provide the necessary baseline upon which to assess the influence of increasing human activities on this endemic, Vulnerable species.
Conference Paper
Full-text available
How animals interact and develop social relationships in face of sociodemographic and ecological pressures is of great interest. New methodologies, in particular Social Network Analysis (SNA), allow us to elucidate these types of questions. However, the different methodologies developed to that end and the speed at which they emerge make their use difficult. Moreover, the lack of communication between the different software developed to provide an answer to the same/different research questions is a source of confusion. The R package ‘Animal Network Toolkit’ (ANT) was developed with the aim of implementing in one package the different social network analysis techniques currently used in the study of animal social networks. Hence, ANT is a toolkit for animal research allowing among other things to: 1) measure global, dyadic and nodal networks metrics; 2) perform data randomization: pre- and post-network (node and link permutations); 3) perform statistical permutation tests. The package is partially coded in C++ for an optimal computing speed. The package gives researchers a workflow from the raw data to the achievement of statistical analyses, allowing for a multilevel approach: from the individual's position and role within the network, to the identification of interactional patterns, and the study of the overall network properties. Furthermore, ANT also provides a guideline on the SNA techniques used: 1) from the appropriate randomization technique according to the data collected; 2) to the choice, the meaning, the limitations and advantages of the network metrics to apply, 3) and the type of statistical tests to run.
Full-text available
Group-living animals rely on efficient transmission of information for optimal exploitation of their habitat. How efficient and resilient a network is depend on its structure, which is a consequence of the social interactions of the individuals that comprise the network. In macaques, network structure differs according to dominance style. Networks of intolerant species are more modular, more centralized, and less connected than those of tolerant ones. Given these structural differences, networks of intolerant species are potentially more vulnerable to fragmentation and decreased information transmission when central individuals disappear. Here we studied network resilience and efficiency in artificial societies of macaques. The networks were produced with an individual-based model that has been shown to reproduce the structural features of networks of tolerant and intolerant macaques. To study network resilience, we deleted either central individuals or individuals at random and studied the effects of these deletions on network cohesiveness and efficiency. The deletion of central individuals had more negative effects than random deletions from the networks of both tolerant and intolerant artificial societies. Central individuals thus appeared to aid in the maintenance of network cohesiveness and efficiency. Further, the networks of both intolerant and tolerant societies appeared to be robust to the loss of individuals, as network fragmentation was never observed. Our results suggest that despite differences in network structure, networks of tolerant and intolerant macaques may be equally resilient.
Full-text available
Since group-living animals are embedded in a network of social interactions, socioecological factors may not only affect individual behavioural strategies but also the patterning of group-level social interactions, i.e. the network structure. These co-variations between socioecological factors, individual behaviour, and group-level structure are important to study since ecological factors may strongly influence animal health outcomes and reproductive success. Besides factors such as social information and/or infectious agents, with far-reaching individual fitness consequences, seem independent of individuals’ own social interactions but directly affected by the topology of the social network. This paper reviews how socio-ecological pressures, i.e., causal factors (food distribution, predation and infectious agent risk), via intermediary mechanisms (stress, information sharing and mating system), may affect individual social behaviour and consequently, social network topology. We also discuss how evolutionary driving forces, genetic (i.e. genes) and cultural (i.e. learned behaviour) selection, may result in a specific composition of individuals’ social strategies that produce network topologies that might be optimized to specific socio-ecological conditions. We conclude that studies focusing on whether and how well networks resist to changing conditions might provide a better understanding of the rules underlying individual behaviour, which in turn influences network topology - a process we have called network evolution. Evolutionary processes may favour a group phenotypic composition, thus a network topology. This has been referred to as a “collective social niche construction”.
Full-text available
Network approaches to ecological questions have been increasingly used, particularly in recent decades. The abstraction of ecological systems – such as communities – through networks of interactions between their components indeed provides a way to summarize this information with single objects. The methodological framework derived from graph theory also provides numerous approaches and measures to analyze these objects and can offer new perspectives on established ecological theories as well as tools to address new challenges. However, prior to using these methods to test ecological hypotheses, it is necessary that we understand, adapt, and use them in ways that both allow us to deliver their full potential and account for their limitations. Here, we attempt to increase the accessibility of network approaches by providing a review of the tools that have been developed so far, with – what we believe to be – their appropriate uses and potential limitations. This is not an exhaustive review of all methods and metrics, but rather, an overview of tools that are robust, informative, and ecologically sound. After providing a brief presentation of species interaction networks and how to build them in order to summarize ecological information of different types, we then classify methods and metrics by the types of ecological questions that they can be used to answer from global to local scales, including methods for hypothesis testing and future perspectives. Specifically, we show how the organization of species interactions in a community yields different network structures (e.g., more or less dense, modular or nested), how different measures can be used to describe and quantify these emerging structures, and how to compare communities based on these differences in structures. Within networks, we illustrate metrics that can be used to describe and compare the functional and dynamic roles of species based on their position in the network and the organization of their interactions as well as associated new methods to test the significance of these results. Lastly, we describe potential fruitful avenues for new methodological developments to address novel ecological questions.
Full-text available
How animals interact and develop social relationships regarding, individual attributes, sociodemographic and ecological pressures is of great interest. New methodologies, in particular Social Network Analysis, allow us to elucidate these types of questions. However, the different methodologies developed to that end and the speed at which they emerge make their use difficult. Moreover, the lack of communication between the different software developed to provide an answer to the same/different research questions is a source of confusion. The R package Animal Network Toolkit (ANT) was developed with the aim of implementing in one package the many different social network analysis techniques currently used in the study of animal social networks. Hence, ANT is a toolkit for animal research allowing among other things to: 1) measure global, dyadic and nodal networks metrics; 2) perform data randomization: pre-network and network (node and link) permutations; 3) perform statistical permutation tests. The package is partially coded in C++ for an optimal coding speed, and it gives researchers a workflow from raw data to the achievement of statistical analyses, allowing for a multilevel approach: from individual position and role within the network, to the identification of interaction patterns, and the analysis of the overall network properties.
Animal social networks are shaped by multiple selection pressures, including the need to ensure efficient communication and functioning while simultaneously limiting disease transmission. Social animals could potentially further reduce epidemic risk by altering their social networks in the presence of pathogens, yet there is currently no evidence for such pathogen-triggered responses. We tested this hypothesis experimentally in the ant Lasius niger using a combination of automated tracking, controlled pathogen exposure, transmission quantification, and temporally explicit simulations. Pathogen exposure induced behavioral changes in both exposed ants and their nestmates, which helped contain the disease by reinforcing key transmission-inhibitory properties of the colony’s contact network. This suggests that social network plasticity in response to pathogens is an effective strategy for mitigating the effects of disease in social groups.
Social substructure can influence pathogen transmission. Modularity measures the degree of social contact within versus between "communities" in a network, with increasing modularity expected to reduce transmission opportunities. We investigated how social substructure scales with network size and disease transmission. Using small-scale primate social networks, we applied seven community detection algorithms to calculate modularity and subgroup cohesion, defined as individuals' interactions within subgroups proportional to the network. We found larger networks were more modular with higher subgroup cohesion, but the association's strength varied by community detection algorithm and substructure measure. These findings highlight the importance of choosing an appropriate community detection algorithm for the question of interest, and if not possible, using multiple algorithms. Disease transmission simulations revealed higher modularity and subgroup cohesion resulted in fewer infections, confirming that social substructure has epidemiological consequences. Increased subdivision in larger networks could reflect constrained time budgets or evolved defences against disease risk. © 2018