Content uploaded by Johann Mourier
Author content
All content in this area was uploaded by Johann Mourier on Oct 07, 2017
Content may be subject to copyright.
Evidence of social communities in a spatially structured network of a free-ranging
shark species
Johann Mourier
*
, Julie Vercelloni, Serge Planes
Laboratoire d’Excellence «CORAIL »and USR 3278 CNRS-EPHE, Centre de Recherche Insulaire et Observatoire de l’Environnement (CRIOBE) and Centre de Biologie et d’Ecologie
Tropicale et Méditerranéenne
article info
Article history:
Received 5 April 2011
Initial acceptance 30 June 2011
Final acceptance 14 October 2011
Available online 19 December 2011
MS. number: 11-00283
Keywords:
association pattern
Carcharhinus melanopterus
community structure
shark
social network
spatial ecology
Large, solitary, marine predators such as sharks have been observed to aggregate at specific areas. Such
aggregations are almost certainly driven by foraging and behavioural strategies making space for diverse
spatial organizations. Reef-associated shark species often show strong patterns of site fidelity that could
be viewed as a prerequisite for sociality. However, there is limited empirical evidence that such aggre-
gations are driven by intrinsic social factors. Association data for blacktip reef sharks, Carcharhinus
melanopterus, were obtained from photoidentification surveys conducted in Moorea coral reefs (French
Polynesia). We adapted a social network approach to demonstrate evidence of four main communities
and two subcommunities within the population. We confronted the resulting structure with candidate
explanatory variables. Sharks formed spatial groups characterized by nonrandom and long-term asso-
ciations, despite opportunities for social relationships to develop between communities. Sex and length
of sharks tended to influence assortment at the population and community levels. Individual space use
also explained community structure, although spatial assortment was globally weaker than random
expectations, suggesting that observed associations were not an artefact of the sampling design or spatial
distribution of individuals. We conclude that the observed grouping patterns not only resulted from
passive aggregations for specific resources, but rather the communities developed from an active choice
of individuals as a sign of sociability. Individual preferences and adaptation to local conditions, as well as
demographic, ecological and anthropogenic factors, may explain the social variability between
communities. This suggests that a stable grouping strategy may confer substantial benefits in this marine
predator.
Ó2011 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Understanding spatial ecology and dynamics is a prerequisite
for the conservation and management of particular species. Spatial
patterns are influenced by many interacting factors that are often
difficult to assess and to disentangle. The movement dynamics
within the habitat and conspecific interactions within a population
are fundamental as they directly influence the genetic structure
and the nature of the population’s habitat usage (Holyoak et al.
2008). Overall, it has been shown that animal spatial use relies on
complex processes driven by individual short-term strategies
maximizing individual benefits (e.g. reproduction, feeding and
survival), together with interactions of nearby conspecifics that are
part of their local environment. Meanwhile, movements are also
likely to be shaped by longer-term fitness implications, such as
avoidance of inbreeding (Holyoak et al. 2008). While the impor-
tance of social behaviour for spatial use has been widely investi-
gated among terrestrial predators (Sandell 1989; Atwood & Weeks
2003; Wagner et al. 2008) and marine mammals (Lusseau et al.
2006; Wiszniewski et al. 2009; Frère et al. 2010), such informa-
tion is still missing in most large, free-ranging, marine fish. There is
a need for additional information regarding the influence of
intraspecific associations on the spatial structuring of populations.
Such information is particularly important in elasmobranchs given
their longevity and vulnerability to overfishing (Stevens et al.
2000).
While shark species are often viewed as solitary hunting
animals, groups and aggregations are relatively common across
phylogeny and ecology in sharks (Springer 1967) and this implies
that grouping is a common form of spatial distribution among
sharks (Jacoby et al. 2011). Grouping behaviours are seen in the
Heterodontiformes (Powter & Gladstone 2009), Hexanchiformes
(Ebert 1991), Squatiniformes (Standora & Nelson 1977) and
Carcharhiniformes (Klimley & Nelson 1981; McKibben & Nelson
*Correspondence: J. Mourier, USR 3278 CNRS-EPHE, Centre de RechercheInsulaire
et Observatoire de l’Environnement (CRIOBE), BP 1013e98 729, Papetoai, Moorea,
French Polynesia and Centre de Biologie et d’Ecologie Tropicale et Méditerranéenne,
Université de Perpignan, 66860 Perpignan, France.
E-mail address: johann.mourier@gmail.com (J. Mourier).
Contents lists available at SciVerse ScienceDirect
Animal Behaviour
journal homepage: www.elsevier.com/locate/anbehav
0003-3472/$38.00 Ó2011 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.anbehav.2011.11.008
Animal Behaviour 83 (2012) 389e401
1986). Grouping has been reported in planktivores (Meekan et al.
2006), large predators (Domeier & Nasby-Lucas 2007), coastal
species (Klimley & Nelson 1981; Heupel & Simpfendorfer 2005) and
reef-associated sharks (Stevens 1984; McKibben & Nelson 1986;
Economakis & Lobel 1998; Speed et al. 2011). Some of the poten-
tial functions of these groupings have been attributed to commu-
nication or transfer of social information (Klimley & Nelson 1981),
courtship (Sims et al. 2000), cooperative hunting (Ebert 1991) and
group protection or avoidance of sexual harassment (Economakis &
Lobel 1998; Wearmouth & Sims 2008). Sharks have also been
observed to form dominance hierarchies (Allee & Dickinson 1954;
Myrberg & Gruber 1974), as well as being capable of learning
(Clark 1959; Guttridge et al. 2009b). In fact, sharks’relative brain
massebody mass ratios were found to be comparable to those of
mammals (Northcutt 1977; Yopak et al. 2007), suggesting that they
are capable of complex social behaviours such as those demon-
strated in mammals and birds (Striedter 2005). However, the highly
mobile nature of sharks, combined with the difficulty of following
individuals in the open sea, has made examination of social inter-
actions or associations difficult. Actually, hypotheses of intraspecific
associations and grouping mostly rely on direct field observations
(Economakis & Lobel 1998) and some recent tracking data such as
acoustic telemetry (Heupel & Simpfendorfer 2005) or proximity
receivers (Holland et al. 2009; Guttridge et al. 2010; Krause et al.
2011). In these surveys, inshore sharks in tropical islands showed
restricted home ranges together with some degree of site attach-
ment (Stevens 1984; McKibben & Nelson 1986; Papastamatiou et al.
2009). Long-term fidelity combined with a high degree of home
range overlap between tracked individuals would fit with the
hypothesis of the existence of persistent associations between
individuals (i.e. social groups of shark). In fact, despite the aggre-
gative nature of some shark species, so far no study has investigated
the influence of conspecific associations on spatial use for any free-
ranging shark species. It remains difficult to determine whether
observed groups of sharks in the wild reflect only aggregative
behaviour or more complex stable social entities. Recent studies of
captive sharks suggested that they were able to form nonrandom
associations, showing an active preference when resources were
controlled for (Guttridge et al. 2009a; Jacoby et al. 2010). In addi-
tion, wild juvenile sharks showed assortative associations within
their nursery (Guttridge et al. 2011). These studies highlight the
need for additional research to investigate the importance of social
factors for space use in the wild.
The blacktip reef shark, Carcharhinus melanopterus, is a common
shark species of Indo-Pacific coral reefs (Compagno et al. 2005).
This species inhabits shallow reefs and sand-flats both in atolls and
high islands (Nelson & Johnson 1980; Stevens 1984; Papastamatiou
et al. 2009, 2010). Although this species is abundant around French
Polynesian islands, many aspects of its basic natural history remain
poorly documented. It is not considered to be either a solitary or
a schooling shark, but is often observed in small aggregations,
especially when feeding (Nelson & Johnson 1980; Papastamatiou
et al. 2009). It also shows a high degree of site attachment and
spatial overlap (Stevens 1984; Papastamatiou et al. 2009). This
suggests that these aggregations and grouping patterns might be
stable over time; however, further studies are needed to test this
relationship. Site attachment and spatial overlap make this species
an ideal model to test for the presence of social organization in free-
ranging reef shark populations and to determine factors affecting
these associations.
Space use and ranging patterns of individuals have commonly
been used to investigate social structure in animal populations. This
is because the amount of spatial overlap between individuals
provides indirect information about the probability of social
interactions (Clutton-Brock 1989). In most animal studies, the
relationship between individuals is defined by time spent together
using an association index (Whitehead 2008). The problem with
this approach is that association patterns based upon the time
spent together can be influenced by both individual ranging
patterns and intrinsic social affiliations (Lusseau et al. 2006). In
many different species, some individuals showed association
patterns that correlated with their home range overlap (Chaverri
et al. 2007; Wolf et al. 2007; Frère et al. 2010). Consequently,
estimates of association patterns may be biased because individuals
with similar ranging patterns are more likely to be sighted together,
even if spatial overlap does not necessarily account for association
patterns (Carter et al. 2009). It thus becomes important to tease
apart aggregative behaviour driven by external forces, such as prey
distribution or habitat preferences, from those driven by intrinsic
social preferences.
Quantifying the structure of an animal society is difficult,
because it represents a complex agglomeration of individuals in
which relationships change in time and space. Recently, social
network theory has offered a powerful set of statistics for charac-
terizing and analysing individual associations within a population-
level social context (Krause et al. 2007, 2009; Sih et al. 2009). These
tools have greatly facilitated our understanding of how ecological,
social and population-level factors influence association patterns
(Croft et al. 2005; Sundaresan et al. 2007; Wolf et al. 2007). Where
social groups are not discretely structured (Lusseau et al. 2006)
recent network techniques (Lusseau et al. 2008) can help detect
statistically significant structure in the population. This technique
appears well adapted to the monitoring of associations of sharks
through space and time.
Despite considerable advances in telemetry and remote tracking
of free-ranging sharks (Sims 2003), there is a need for specific
research into the major factors determining social structure of
marine predators (Wearmouth & Sims 2008), especially in the
context of currently heavy exploitation of sharks (Baum et al. 2003).
Our present study used an original approach to analyse the spatial
dynamics of a reef shark species exhibiting aggregative behaviour.
It aimed to determine whether community structure was present in
the studied population using recently developed network tech-
niques. We define a ‘community’as a set of individuals that are
more associated among themselves than they are to the rest of the
population (Croft et al. 2008). Furthermore, if such structure is
detected in the population, we can expect assortment by external
factors relating to habitat and space use and by internal factors such
as sex and age (or length; Wolf et al. 2007), or individual prefer-
ences for interaction (Pomeroy et al. 2005). Indeed, size or sex
assortment is relatively common in reef shark groups (Sims 2003;
Wearmouth & Sims 2008; Jacoby et al. 2011). We used photo-
identification techniques to describe association patterns in a pop-
ulation of 133 sharks along the reef. We then tried to determine
whether potential structure in associations reflects solely an
aggregative behaviour governed byextrinsic factors, such as habitat
preferences, or underlies more complex social preferences.
METHODS
Study Sites
The study was conducted at Moorea Island (17
30
0
S; 149
51
0
W)
in the Society archipelago, French Polynesia. A total of seven sites
were surveyed on a regular basis along 10 km of the north shore of
Moorea (Fig. 1). Sites were selected for various reasons.
(1) Moorea council implemented a Management Plan for Marine
Environment (Plan de Gestion de l’Espace MaritimeePGEM) in
October 2004 that includes two areas for shark-feeding activities on
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401390
the outer reef that we selected as Sites 2 and 6 (Fig. 1) and where
provisioning occurs daily between 0800 and 1000 hours.
(2) Inside the lagoon of Moorea, the pink whipray, Himantura fai,
has an active feeding site (Site 5, Fig. 1) that consistently attracts
blacktip reef sharks (Gaspar et al. 2008).
(3) Shark aggregations are also observed in another recreational
diving site on the outer reef (Site 4, Fig. 1) where recreational shark
feeding was stopped about 7 years ago.
(4) Finally, three additional sites, where feeding does not occur,
were selected for surveys on the outer slope (Sites 1, 3 and 7, Fig. 1).
Sites 1, 2, 3, 4, 6 and 7 were located on the outer reef and were
characterized by coral structures from the barrier reef to the drop
out (70 m depth). Site 5 was located inside the lagoon between 2
and 10 m depth within a small channel and was characterized by
coral patches in a sandy habitat.
Shark Identification and Data Collection
Shark photoidentification has already been used in many
species, mainly large-bodied ones, such as Carcharodon carcharias
(Domeier & Nasby-Lucas 2007) and Rhincodon typus (Meekan et al.
2006), using natural variations in colour patterns on the body. It has
already been used on the blacktip reef shark (Porcher 2005) to look
at the specific shape of margins separating black, white and brown
colour patterns on the dorsal fin. The succession of coloured lines
was shown to vary consistently and to be unique to each individual
(Porcher 2005). In the present study, we used photoidentification of
both sides of the dorsal fin(Fig. 2) as well as other distinctive marks
such as scars, notches and dots throughout the shark’s body.
To study spatial and temporal overlap between sharks, we
implemented underwater surveys consisting of dives lasting less
than 1 h (mean dive duration SE ¼49.46 0.39 min) at a depth of
about 15 m on the outer reef of Moorea and 5 m inside the lagoon
(Site 5) using a stationary-point technique (Ward-Paige et al. 2010).
Throughout the study a trained diver recorded and photographed
sharks observed within about 15e20 m of the diver. A second diver
was present for safety reasons but did not get involved in the
monitoring. The observer recorded the sex of individuals by the
presence or absence of claspers. Total length (TL) was measured
from the specimen photographed fromthe side. As a result, 66% of all
sharks of the network (see below) were measured. These known-
sized sharks were then used as visual markers for estimating the
sizes of other sharks present inside the network. Shark length was
classified into size classes ranging from 1 to 6 (1: TL <110 cm;
2: 110e119 cm; 3: 120e129 cm; 4: 130e
139 cm; 5: 140e150 cm;
6: TL >150 cm). Although Papastamatiou et al. (2009) found that
male blacktip reef sharks reach maturity at about 100 cm TL, size at
maturity in Moorea was 110 cm (J. Mourier, unpublished data).
Every new shark was recorded on a specific identification sheet,
similar to the work done on the sicklefin lemon shark, Negaprion
acutidens, in Moorea (Buray et al. 2009). Identification was facili-
tated by the good visibility of Moorea waters, being relativelystable
over time and allowing photography of some shy specimens that
remained up to 20 m from the diver. Our surveycumulated 190 dives
conducted between February 2008 and June 2010 (i.e. 21 on Site 1,
46 on Site 2, 9 on Site 3, 33 on Site 4, 40 on Site 5, 31 on Site 6 and 10
on Site 7; Fig. 1). Only one dive was conducted at asingle site on the
same day. Dives were conducted in the afternoon outside of provi-
sioning hours.
Network Analysis
Defining associations and network construction
For the majority of species, social interactions are difficult to
observe directly, especially underwater; they might occur out of
sight or infrequently. In this case, the usual approach is to infer
social relationships between individuals based on accumulated
observations of social associations (i.e. based on group composi-
tion, nearest neighbours or spatial proximity). When using group-
based data, we involved the ‘gambit of the group’(Whitehead &
Dufault 1999), assuming that behavioural interactions occur
within groups and repeated group membership is an indicator of
the strength or frequency of these interactions. For species such as
sharks, in which interactions are hard to describe, we define
association as the simultaneous occurrence of two or more indi-
viduals at the same site (Whitehead 2008). Thus, associations were
based on ‘co-occurrence’, such that individuals present during the
same dive and within the diver’s visual range (i.e. 15e20 m radius)
were considered as part of the same group (referred to hereafter as
‘shark groups’). Here we describe associations between individuals
Figure 1. Location of study sites on the north coast of Moorea. Shark feeding occurs in Sites 2 and 6, and ray feeding in Site 5.
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401 391
that may be passively sharing time and space. However, during dive
surveys, sharks were regularly involved in interactions, which are
more complex behaviours that are typically directed, such as nose
to tail following, parallel swimming or circling (Myrberg & Gruber
1974; Guttridge et al. 2011), and loose aggregations (see Appendix
for illustration). By cumulating the co-occurrences over a series of
underwater surveys of the population at different sites, we can
build an association matrix between pairs of sharks. Strength of
association among pairs of individuals was calculated using the
half-weight index (HWI; Cairns & Schwager 1987) in SOCPROG 2.4
(Whitehead 2009) and was restricted to individuals sighted at least
five times over the entire study period (greater than or equal to the
median number of sightings per individuals; median ¼5,
mean 95% confidence interval, CI ¼8.38 1.03; range 1e39) as it
is commonly used for accurate descriptions of associations (Lusseau
et al. 2006; Wiszniewski et al. 2009, 2010). The sampling period in
the analysis was set to 15 days to accommodate the rate of data
collection at different sites.
Randomization techniques
To quantify population structure, it is necessary first to establish
that the observed data provide statistical evidence that the pop-
ulation contains nonrandom structure (Whitehead et al. 2005). To
understand the importance of social behaviour in the observed
association data, it is necessary to disentangle the contributions of
social preferences, gregariousness and sampling to the observed
association indexes. We can compare the real data to that produced
by making associations random to determine whether individuals
display nonrandom structure in the studied population. However,
randomization is not trivial in networks (Croft et al. 2011). We used
a modified version of the BejdereManly method, which is used to
randomize association data to obtain null random networks that
control for sampling structure and gregariousness of individuals
(Manly 1997; Bejder et al. 1998; Whitehead et al. 2005). We
permuted group membership so that group size and the number of
groups in which each individual was identified were both the same
as in the original data set. We did this by a series of flips in which
randomly chosen records of individual A in group 3 and individual
B in group 7, for example, were flipped to A in 7 and B in 3 (Manly
1997). Therefore, to determine whether associations in the studied
population were significantly different from random, the original
association matrix was randomized 1000 times with 100 flips per
permutation within sampling periods (Whitehead et al. 2005).
A significantly higher coefficient of variation (CV) of real association
indices compared to that of randomly permuted data indicates the
presence of long-term preferred companions in the population
(Whitehead 1999). The randomization procedures were computed
in R version 2.11.1 (The R Foundation for Statistical Computing,
Vienna, Austria, http://www.r-project.org).
Community structure
Many methods for detecting communities within social
networks have been described in recent years (Whitehead 2008).
The blacktip reef shark community structure in our studied area
was examined using the modularity matrix clustering technique
described by Newman (2006) and Lusseau et al. (2008). The
modularity matrix is the association index (i.e. weight) between
two individuals minus the expected weight if associations are
randomly distributed in the population. The eigenvector of the
dominant eigenvalue of the modularity matrix is then used to split
the matrix successively into two clusters. This divisive procedure is
then iterated on all resulting clusters. The most parsimonious
division in the network is subsequently determined by the division
that maximizes the modularity coefficient, Q(Lusseau et al. 2008).
Recently, a measure of uncertainty was introduced in this proce-
dure by Lusseau et al. (2008) and increases significantly the accu-
racy of defining real communities within the global network
(Wiszniewski et al. 2010). To assess the degree of confidence for the
communities identified, we bootstrapped observed group
membership samples (shark groups sampled during the dives)
1000 times by resampling (with replacement) these samples. The
replicates were obtained using 15-day sampling periods with the
same sample size as real data. We then applied the modularity
community division algorithm described above on each bootstrap
Figure 2. Photoidentification of blacktip reef sharks. (a, b) Global view of both sides of (a) a female ‘Op19’and (b) a male ‘V19M’; note the elongated claspers that extend beyond the
pelvic fins in males (b) and their absence in females (a). (ceh) Photographs of both sides of the dorsal fin of six individual C. melanopterus: (c) ‘V12M’; (d) ‘Tao25’; (e) ‘Op26’;
(f) ‘Op18’; (g) ‘Op27’and (h) ‘V21M’taken between 2008 and 2010; note the margin patterns between the black and white parts of the dorsal fin, which are unique to each
individual. (i, j) Persistence of patterns over 10 years for (i) ‘V27F’and (j) ‘V12F’. Photos: (aeh) Johann Mourier; (i, j top) Ila Porcher.
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401392
replicate (Lusseau et al. 2008; Wiszniewski et al. 2010).
A comembership matrix of the proportion of times that two sharks
were clustered in the same group over all bootstrap community
division replicates was built. The bootstrap procedure was
computed in R version 2.11.1. We subsequently carried out the
modularity community division algorithm on this comembership
matrix (Lusseau et al. 2008). The comembership matrix and
resulting community structure (determined by Q
max
) were visual-
ized with NETDRAW (Borgatti et al. 2002). The ‘spring embedding’
algorithm with node repulsion was used for laying out the nodes’
positions (Borgatti et al. 2002). The spring embedding algorithm
achieves a layout of the network with densely connected nodes
clustered together and nodes with few connections placed around
the edge (Croft et al. 2008). Thus groups of well-connected indi-
viduals tend to be grouped together in the resulting visualization.
We also tested the significance of Qbyusing a randomization test.
Following the randomization technique described above, we applied
the modularity matrix clustering technique to find communities
within 1000 random networks. We used the maximum Qas the
statistic test.If the observed data gave riseto a Qva lue in the top 5 % of
the randomized values, we rejected the hypothesis that the observed
value could have arisen by chance alone.
We then used lagged association rate (LAR) techniques imple-
mented in SOCPROG (Whitehead 1995) to compare the temporal
stability of associations within and between these social groups.
Such a combined approach has the advantage of distinguishing
temporally stable social groups from the short-term clustering of
individuals (Whitehead 2008). Each LAR was compared to the null
association rate, which is the expected LAR in the absence of any
preferred associations. The precision of the LARs was estimated
using a jackknifing over 30-day periods (Whitehead 1995).
Spatial patterns
We then describe spatial patterns through a histogram of
percentage of sightings of each individual at each site during our
190 dives standardized for sampling effort per site. We generated
aBrayeCurtis similarity matrix and used an ANOSIM test to
compare space use among sharks of each community found in the
previous analysis. We used the percentage of sightings for each
shark at each studied site as the dependent variable, set community
membership as a factor in the model and then used a one-way
ANOSIM test to detect pairwise differences. The ANOSIM test
compares calculated overlap values against simulated overlap
produced from 999 random permutations. Statistical significance
indicates that a pair of communities had low spatial overlap.
ANOSIM also generates a global Rvalue, ranging from 1to1,
corresponding to the degree of similarity between communities.
We used Clarke & Warwick’s (2001) criterion where R<0.25
indicates high overlap between groups, Rof 0.25e0.75 indicates
moderate separation between groups, and R>0.75 indicates a high
degree of spatial segregation between groups. We used a nonmetric
multidimensional scaling ordination (nMDS) to draw community
differences in space. ANOSIM and nMDS analyses were performed
using PRIMER version 5.0 (http://www.primer-e.com).
Assortment patterns
Assortment in a network describes the tendency of individuals
to be connected to any other individuals that share some charac-
teristics. We considered assortment by sex (male, female), length
difference and spatial overlap. We performed correlation analyses
between the HWI matrix and matrices for either pairwise length
difference or sex similarity. The length difference matrix was based
on the pairwise difference of length classes from 1 to 6 as defined
above and values ranged from 0 (same length class) to 5 (5 length
classes difference). We compared the observed Pearson correlation
coefficient, r, with 1000 coefficients derived from randomized
networks using the procedure described above. Pvalues for each
comparison represent the proportion of correlation coefficients
from randomized networks that were greater in magnitude
(depending on the direction of the correlation) than the observed
correlation coefficient. We also used the correlation coefficient
between the matrix of association indices and the BrayeCurtis
similarity matrix for spatial overlap following the same procedure.
These analyses were performed for both the global network and
each community previously identified.
Sex ratios for C. melanopterus were calculated for the whole
network and then by community. In each case, sex ratios were
compared against an expected 1:1 ratio using a chi-square test for
goodness of fit.
Sociality at the community level
To investigate variability in shark’s sociality, we permuted
groups of sharks 1000 times, keeping constant the number of
individuals in each group and the number of groups in which each
shark was observed (Whitehead 2008). The standard deviation of
the typical group size (TGS), that is the size of a group as experi-
enced by an individual (Jarman 1974), was used to identify sharks
consistently found in larger or smaller groups.
We then calculated five egocentric network measures (strength,
eigenvector centrality, reach, clustering coefficient and affinity) to
investigate differences in the centrality of individuals and groups.
These measures were calculated from the weighted association
matrix (HWI) in SOCPROG (Whitehead 2009). Strength is a measure
of gregariousness and is the sum of the association indices for each
individual (i.e. weighted degree); eigenvector centrality indicates
the level of centrality of an individual to its associates as well as the
centrality of its associates, and is a measure of how well connected
an individual is; reach is a measure of indirect connectedness; the
clustering coefficient indicates the tendency for a focal individual’s
associates to be associates themselves; and affinity is a measure of
the strength of an individual’s associates, weighted by the associ-
ation index between them, and determines whether individuals
connect strongly to individuals that also have high strength.
Following the same randomization procedure, we compared
centrality measures to those of 1000 random networks to test
whether network properties were influenced by individual asso-
ciation preferences (Lusseau et al. 2008). In addition, we tested
whether sex could explain the variation observed among individ-
uals in centrality level as well as the differences in the centrality of
the communities. We then compared the mean network metrics
between communities using a randomization test implemented in
the software PERM (Duchesne et al. 2006). We estimated Pvalues
by comparing the observed mean value to a null distribution of
values generated from 10 iterations of 1000 random permutations
of the data, in which the observed number and size of communities
were kept constant (Duchesne et al. 2006). We applied sequential
Bonferroni corrections to each pairwise comparison (Rice 1989).
RESULTS
Underwater Surveys
Underwater surveys showed group size ranged from two to 29
individuals per dive (mean SE ¼10.65 0.38 individuals). Site-
specific group sizes (mean SE) were 7.33 0.88 (Site 1),
14. 93 0.79 (Site 2), 5.77 0.66 (Site 3), 9.30 0.43 (Site 4),
8.97 0.72 (Site 5), 13.61 0.68 (Site 6) and 4.60 0.67 (Site 7). Out
of 241 individuals (150 males, 91 females) identified underwater
over the 190 dives and across the seven study sites, 133 individuals
(88 males, 45 females) were observed more than four times and
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401 393
AA1 A2 BCD
N = 54
N = 54
N = 29
N = 29
N = 25
N = 25
N = 14
N = 14
N = 30
N = 30
N = 35
>150
140-149
130-139
120-129
110-119
<110
0
20
40
0
20
40
0
20
Frequency (%)
Linear projection of sites Total length (cm)
0
40
80
0
40
80
0
40
80
0
40
80
0
40
80
0
40
80
1000
700
100
400
Mean percentage site use
40
0
20
40
0
20
20
40
40
0
N = 35
(b) (c)
Number of bootstrap replicates
in which the two individuals
were members of the same
community:
(a)
1234567
Figure 3. (a) The shark social network based on bootstrap replicates of the association matrix. Each individual in the network is represented by a node. Community structure was
identified using the modularity matrix technique and membership in the five communities is displayed by colour (dark grey ¼community A (A1 þA2); dark blue ¼subcommunity
A1; light blue ¼subcommunity A2; green ¼community B; yellow ¼community C; red ¼community D). Communities were composed of both males (squares) and females
(circles). The shade of the edges represents the likelihood of individuals belonging to the same group (the darker, the greater). Node size is proportional to the length of individuals.
Note that the placement of nodes represents social, rather than spatial, proximity, although the two may often be correlated. (b) Spatial profiles of communities from the overall
network. Numbers on the abscissa correspond to the sites described in Fig. 1. Values on the ordinate indicate the mean percentage use of each site by members of each community.
Error bars indicate SE. (c) Total length frequency distributions for each community.
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401394
were retained for the network analysis (Fig. 3a). The smallest indi-
vidual was about 100 cm in total length (Fig. 3c) and no neonates
were present within the population. Significantly higher mean CV of
association indices (CV ¼1.849; random CV ¼1.19 3; P<0.001)
indicates that long-term preferred and avoided companions are
present in the population (Whitehead et al. 2005).
Community Structure
Applying the modularity matrix technique on bootstrap
replicates of the association matrix divided the population of 133
sharks into four main social communities (Q
max
¼0.62, P<0.001;
Fig. 3a). Since Q
max
was high, the analysis was repeated within each
community to investigate fine-scale structure within a community.
This resulted in a division of the A community into two subcom-
munities (A1 and A2; Q
max
¼0.48; Fig. 3a). Based on bootstrap
replicates, it appears, however, that there is less confidence in the
community membership of A (Fig. 3a) suggesting that associations
within A may change more frequently. Mixed-community groups
were sighted 32.1% of the time. During the breeding season
(NovembereMarch; Porcher 2005), there were 36.6% of mixed-
community groups, while they occurred 29.4% of the time during
the nonbreeding season (AprileOctober). Mean associations were
significantly lower between than within communities (Mantel test:
P<0.001; Table 1). Together, association patterns were signifi-
cantly higher within than between sexes (Mantel test: P<0.01;
Table 1). Lagged association rate analysis confirmed the segregation
of the communities, as associations within each of the communities
were temporally stable and remained above the null association
rate over the entire study period (Fig. 4a). In contrast, intercom-
munity LARs were relatively lower and stayed close to the null
association rate, except between communities A and B meaning
that intercommunity associations were temporally unstable
(Fig. 4b). Relatively high intercommunity LAR between A and B
suggests that greater flow of associative ties may exist between
members of these communities (Fig. 4b).
Space Use and Assortment Patterns
Sharks were sighted an average of 89.48% of the time at a single
site (range 43e100%) and all communities were almost restricted to
a unique site, with only a few individuals occasionally visiting
another site (Fig. 3b). There were significant differences in space
utilization patterns between communities with significant spatial
separation (low overlap; ANOSIM: R¼0.866, P<0.001 when
including the two subcommunities A1 and A2), although there was
high overlap for clusters A1 and A2 (R¼0.041, P¼0.047), which
was apparent in the nMDS plots (Fig. 5). At population (i.e. global
network), community and subcommunity levels, sharks were
positively spatially assorted, with Pearson correlation coefficients
being strong and positive although assortments were significantly
weaker than random expectations for the global network as well as
for communities A and D (P<0.05; Table 2) but stronger for
community C (P<0.05; Table 2) and not different from random
expectations for community B (Table 2). Moreover, preferred
associations, defined by pairwise association estimates at or above
the 97.5% percentile (i.e. HWI >0.56), were found between sharks
showing as little as 45% spatial overlap, while some pairs showed
temporal avoidance (i.e. HWI ¼0) despite 100% spatial overlap in
the case of subcommunities A1 and A2.
At the population, community and subcommunity levels, sharks
were weakly but significantly assorted by sex and length. Pearson
correlation coefficients were positive and significantly stronger than
random expectations for matrices of HWI and sex similarity and
were negative but significantly stronger than random expectations
for matrices of HWI and length distance as well as sex and length
combined (P<0.001 for all comparisons; Tabl e 2). Distributions of
total length of individuals were different between communities,
with community C having a significantly greater proportion of longer
individuals than community B (KolmogoroveSmirnov test: P<0.01)
Table 1
Mean association index for each sex and each community
Classes Within classes NMean
association
Maximum
association
Community All 133 0.09 (0.03) 0.62 (0.13)
Community A 54 0.24 (0.09) 0.60 (0.13)
Community A1 29 0.27 (0.12) 0.59 (0.17)
Community A2 25 0.21 (0.06) 0.52 (0.10)
Community B 14 0.35 (0.12) 0.62 (0.19)
Community C 30 0.28 (0.06) 0.60 (0.07)
Community D 35 0.31 (0.13) 0.65 (0.17)
Within communities 0.28 (0.11) 0.62 (0.14)
Between communities 0.01 (0.02) 0.24 (0.15)
Sex Female 45 0.09 (0.03) 0.59 (0.11)
Male 88 0.09 (0.04) 0.60 (0.15)
Within sex 0.09 (0.04) 0.59 (0.14)
Between sex 0.08 (0.04) 0.54 (0.16)
SEs are reported in parentheses.
10
0.7
(b)
(a) Null
Community A
Community B
Community C
Community D
Null
Community A to B
Community A to D
Community B to C
Community C to D
0.6
0.5
0.4
0.3
0.2
0.1
Lagged association rate
0
0.5
0.4
0.3
0.2
0.1
0
20 30 40 50 60
10 20
La
g
(2 weeks)
30 40 50 60
Figure 4. Lagged association rate for (a) within and (b) between the communities
identified in the network. The null association rate was estimated from all individuals
using 1000 permutations. Standard errors of the LARs were computed by jackknifing
over 30-day periods.
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401 395
and community D (KolmogoroveSmirnov test: P<0.05; Fig. 3c). Sex
ratio at the entire network scale was male biased and differed
significantly from unity (Table 2). Sex ratios within communities
were male biased and significantly different from unity for
community A but not for communities B, C and D (Table 2). Sex
assortment coefficients seemed higher in communities where sex
ratios were balanced (Table 2).
Sociality at the Community Level
Sharks in the studied area differed significantly in their level of
gregariousness (observed SD of TGS ¼2.95; random SD of
TGS ¼2.12; P<0.001). Social behaviour appeared to vary accord-
ing to the sex of the individual (Table 3). Males demonstrated
higher reach and clustering than expected in random networks,
while females showed higher strength and clustering. However,
there was no difference between male and female centrality
measures (ManneWhitney Utest: P>0.05 for all comparisons).
Individual variability in centrality measures appeared to be related
to community membership. Differences in social behaviour
between the communities and differences from random expecta-
tions were strongly supported by network measures (Table 3).
Sharks from community A demonstrated higher eigenvector
centrality, reach, clustering and affinity than expected from
random expectations. Individuals from community C showed
higher strength while individuals from community D had lower
eigenvector centrality but higher clustering than random expec-
tations. Sharks from community A had significantly higher
strength, eigenvector centrality, reach and affinity than all other
communities (P<0.05 after Bonferroni corrections; Table 3),
except when compared to strength of community D (P>0.05).
Furthermore, community D had a significantly higher clustering
coefficient than all other communities (P<0.05 after Bonferroni
corrections for all comparisons; Table 3). Sharks from community
A had a lower mean association rate (Table 1), a lower level of
clustering and higher measures of strength, reach and affinity
(Table 3)indicating that they are the most diversely connected in
the network and are likely to change associates more regularly. A
denser network structure was found in community D in which
individuals are well connected (higher clustering and high mean
association rate).
DISCUSSION
This study revealed a complex structure of associations within
an island shark population initially viewed as a global entity. In
fact, the blacktip reef shark population in our studied area (only
10 km of coastline out of the 60 km shoreline of Moorea Island)
was structured into four communities, with one splitting into two
subcommunities. Communities diverged in their ranging patterns
with individual members forming nonrandom and temporally
stable associations similar to structures observed in classic social
species such as guppies, Poecilia reticulata (Croft et al. 2005),
dolphins, Tursiops spp. (Lusseau et al. 2006; Wiszniewski et al.
2009) or Galápagos sealions, Zalophus wollebaeki (Wolf et al.
2007). The subdivision of the population was found via an
unbiased search within the network of community-level associ-
ations between individuals, derived from association data alone,
without involving any additional information. We then con-
fronted the association patterns with candidate explanatory
variables by testing for the influence of sex, difference in total
length and space use and found that sex and age (i.e. total length)
tended to influence assortment at the population and community
levels and that individual space use patterns also explained
community structure, although spatial assortment of individuals
was globally weaker than random expectations. Overall, high
variability in sociality and organization was found at the
community level.
Individuals form groups that are either ephemeral aggregations
or groups that are highly structured (social groups). Unlike aggre-
gations that are formed by chance because of attraction to a specific
location or a common resource (e.g. food), social groups contain
structure that enables individuals to gain benefits from other
individuals within the group such as foraging efficiency or
protection against predators (Krause & Ruxton 2002) and members
actively seek out specific individuals with which to interact or
group. Determining whether a grouping pattern is an aggregation
Table 2
Degrees of assortment
Assortment level Sex ratio (F:M) PSex Length Sex & length Spatial overlap
Overall social network (OSN) (45:88) <0.001 0.08 (0.03; <0.001) 0.12 (0.06; <0.001) 0.08 (0.04; <0.001) 0.64 (0.73; <0.001)
Community A (13:41) <0.01 0.08 (0.02; <0.001) 0.15 (0.00; <0.001) 0.11 (0.01; <0.001) 0.37 (0.42; <0.001)
Subcommunity A1 (8:21) <0.05 0.10 (0.08; <0.001) 0.23 (0.00; <0.001) 0.20 (0.04; <0.001) 0.35 (0.38; 0.036)
Subcommunity A2 (5:20) <0.01 0.14 (0.00; <0.001) 0.24 (0.03; <0.001) 0.16 (0.03; <0.001) 0.39 (0.41; 0.248)
Community B (6:8) >0.05 0.26 (0.06; <0.001) 0.28 (0.08; <0.001) 0.19 (0.06; <0.001) 0.30 (0.25; 0.061)
Community C (14:16) >0.05 0.26 (0.15; <0.001) 0.30 (0.13; <0.001) 0.16 (0.05; <0.001) 0.22 (0.10; <0.001)
Community D (12:23) >0.05 0.14 (0.01; <0.001) 0.15 (0.03; <0.001) 0.07 (0.03; <0.001) 0.54 (0.57; 0.008)
Sex ratios with Pvalues for chi-square tests are given. Pearson correlation coefficients between association matrix (HWI) and matrices of sex similarity, length distance and
spatial overlap compared with expected coefficients from 1000 random networks together with two-tailed Pvalues are given in parentheses.
Communities
Stress: 0.05
A1
A2
B
C
D
Figure 5. Nonmetric multidimensional scaling ordination of space use between sharks
of the communities. The black dotted line represents community A (A1 þA2).
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401396
or a social group is critical to understanding the evolution of
grouping and possible cost and benefits. In the following, we
discuss the influence of our candidate factors on the grouping
patterns observed here.
Assortment in a Shark Population
Sex and length, taken separately and both combined, were
significant factors explaining community structure within the
blacktip reef shark population of Moorea. Although assortment
coefficients were relatively weak, total length and sex assort-
ments deviated from random expectations at both global pop-
ulation and community levels and were stronger at community
levels. Sex assortment was stronger in communities B and C.
Shark social groups and aggregations made of individuals of the
same sex have been previously documented for numerous species
(Economakis & Lobel 1998; Wearmouth & Sims 2008). The
community C is the only one with members inhabiting primarily
an area restricted to the lagoon (Figs 1,3b) in contrast to all
others inhabiting the fore reef. This community has the particu-
larity of showing stronger sex assortment and a relatively
balanced sex ratio compared to other communities in which
males dominate. Sex segregation in sharks was previously
attributed to sexual dimorphism and to differences in energy
requirements of females to increase reproductive output (Sims
2003). However, in the present population, mixed-sex commu-
nities are found in every habitat, suggesting that sexual segre-
gation is not an exclusive component of this system. Rather, sex
may have controlled the emergence of preferred associations
within mixed-sex communities.
Body length assortment in groups is common among teleost
fishes (Croft et al. 2009) and is known to confer antipredator and/or
foraging advantages (Krause 1994). In a recent study, juvenile
lemon sharks, Negaprion brevirostris, showed size-assortative
associations that were hypothesized to confer a benefit for pred-
ator avoidance (Guttridge et al. 2011). Blacktip reef sharks share
space with potential predators such as the grey reef shark, Carch-
arhinus amblyrhynchos, and the sicklefin lemon shark on the outer
reef of Moorea, which could favour size-assortative grouping
patterns for defence purposes. However, absence of such predators
inside the lagoon (Gaspar et al. 2008; Clua et al. 2010) where
blacktip reef sharks tend to assort strongly by body length does not
support this hypothesis. It is widely recognized that many shark
species show ontogenetic and sex-based shifts in habitat use and
diet composition as juveniles live in nursery habitats and adults in
open habitat (reviewed in Wetherbee & Cortes 2004) and these
passive sorting mechanisms may contribute to the formation of
size- and sex-segregated groups (Sims 2003; Heupel &
Simpfendorfer 2005; Wearmouth & Sims 2008; Jacoby et al.
2011). While communities differed in their demographic compo-
sition (sex and length frequencies), their members tended to
associate preferentially with individuals of similar sex and/or size.
This case study illustrates how assortative associations based on
sex or length may globally influence and structure a shark social
network. As mechanisms producing such preferences remain
difficult to determine precisely, mating strategies, familiarity
developed within cohorts, and size- and sex-dependent dominance
hierarchies (Myrberg & Gruber 1974) could explain such
assortment.
Individual Space Use and Boundaries of Communities
The ranges of individual blacktip reef sharks were almost
restricted to a single site (mean 89% of sightings at one site). Other
studies found a similar degree of site attachment (Stevens 1984;
Papastamatiou et al. 2009, 2010). The advantage of maintaining
a home range has been debated; it may facilitate the use of local
resources such as feeding sites, predator refuges and breeding sites
(Powter & Gladstone 2009; Speed et al. 2011). Fine-scale site
fidelity of gregarious animals is likely to play a key role for a species’
social structure by creating an environment for social relationships
to develop from repeated interactions (Wolf et al. 2007). This may
also benefit individuals by reducing aggression.
As found in other social animals (e.g. Lusseau et al. 20 06; Wolf
et al. 2007), communities in the present study were almost
entirely explained by space segregation with little overlap
Table 3
Average network measures calculated using association strength (HWI) for all individuals and for each community and sex compared to 1000 random networks
Community and
sex class means
Strength Eigenvector centrality Reach Clustering coefficient Affinity
A(N¼54) 13.52 (4.76) 0.12 (0.05)*** 202.47 (75.44)*** 0.30 (0.05)*** 14.80 (1.13)***
Random 13.69 (4.86) 0.10 (0.04) 191.19 (69.76) 0.21 (0.02) 13.86 (0.70)
A1 (N¼29) 13.92 (5.79) 0.13 (0.06)*** 211.56 (90.63)*** 0.31 (0.04) 14.95 (1.13)**
Random 14.00 (5.84) 0.11 (0.05) 205.78 (88.37) 0.25 (0.04) 14.48 (0.93)
A2 (N¼25) 13.05 (3.26) 0.12 (0.04)*** 191.92 (52.62)*** 0.29 (0.06)*14.62 (1.13)***
Random 13.13 (3.30) 0.10 (0.03) 188.25 (50.13) 0.24 (0.04) 14.28 (0.83)
B(N¼14) 7.82 (2.60) 0.02 (0.01) 80.84 (27.00) 0.29 (0.03) 10.32 (0.50)
Random 7.87 (2.56) 0.04 (0.01) 89.50 (30.04) 0.18 (0.02) 11.34 (0.55)
C(N¼30) 9.20 (2.08)** 0.01 (0.00) 89.76 (21.56) 0.27 (0.05) 9.74 (0.32)
Random 8.98 (2.06) 0.03 (0.01) 94.87 (23.59) 0.19 (0.03) 10.54 (0.52)
D(N¼35) 11.97 (4.31) 0.02 (0.02)*** 159.05 (59.78) 0.38 (0.10)*13.08 (1.06)
Random 11.96 (4.33) 0.07 (0.03) 159.36 (59.37) 0.26 (0.06) 13.18 (0.77)
F(N¼45) 11.16 (3.90)*0.05 (0.06) 142.22 (69.91) 0.32 (0.07)** 12.28 (2.25)
Random 11.11 (3.96) 0.07 (0.04) 144.43 (65.05) 0.22 (0.05) 12.67 (1.61)
M(N¼88) 11.73 (4.73) 0.07 (0.06) 158.23 (79.49)*** 0.31 (0.08)** 12.96 (2.29)
Random 11.73 (4.81) 0.08 (0.04) 157.57 (74.92) 0.22 (0.05) 13.06 (1.61)
Overall means 11.54 (4.46)** 0.06 (0.06) 152.81 (76.50) 0.31 (0.08)** 12.73 (2.30)
Random 11.52 (4.52) 0.07 (0.04) 153.01 (72.72) 0.23 (0.05) 12.88 (1.74)
SEs are reported in parentheses. Bold indicates significant differences from 1000 random networks.
*P<0.05; **P<0.01; ***P<0.001.
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401 397
between them. While spatial ranges of communities were
significantly different with low overlap, overall ranges of
communities were not mutually exclusive and mixed-community
groups were observed 32% of the time. Therefore, opportunities
for social relationships to develop between communities of
sharks were present. However, association patterns as well as
LARs were low between communities, suggesting intercommu-
nity associations resulting primarily from aggregative behaviour
(i.e. nonsocial forces such as localized food resources, mating or
predation avoidance, Whitehead 2008). Additionally, the higher
occurrence of mixed-community groups (36.6% of groups) mainly
during the breeding season indicates that mating behaviour is
the major driver influencing association patterns between
communities.
Generally, association patterns among individuals correlate
with spatial overlap (Chaverri et al. 2007). Indeed, if movements of
some animals are restricted in space (e.g. confined to a unique
site), these individuals are more likely to be found in the same
group just by chance. Consequently, given that associating sharks
must overlap in space to some degree, it is not surprising that the
spatial overlap and shark associations were found to correlate
(Table 2). However, other findings do not support this. Although
sharks were positively assorted by their spatial overlap at both the
population and community level, this assortment was globally
weaker than random expectations, suggesting that community
structure resulted from social affiliation and was not an artefact of
the spatial distribution of their members or the sampling proce-
dure. Moreover, preferred associations as defined by pairwise
association estimates at or above the 97.5% percentile (i.e.
HWI >0.56) were found between sharks showing as little as 45%
spatial overlap, while some pairs showed temporal avoidance (i.e.
HWI ¼0) despite 100% spatial overlap. There were also more
study sites than communities demonstrating that site boundaries
are not defining the community boundaries. Finally, spatial
segregation does not exclusively explain the social separation, as
subcommunities A1 and A2 were split from community A with
a high modularity (Q
max
¼0.48) and showed a high degree of
spatial overlap (ANOSIM: R¼0.041). The subcommunity structure
found within community A may be an example of the social
structure containing preferred social associations going beyond
fine-scale site fidelity. Consequently, spatial overlap does not
exclusively explain association patterns, which indicates that the
structure of this population has been driven by active choices of
individuals comparable to complex social structures such as those
observed in dolphins (Lusseau et al. 2006; Wiszniewski et al.
2009; Frère et al. 2010).
Range overlap and the presence of few social ties between
communities also indicate that these sharks tolerated the sporadic
presence of ‘noncommunity members’within the area of their
home range that were not aggressively excluded. Similar spatial
communities have also been described in terrestrial predators
(Macdonald 1983; Bekoff et al. 1984; Gittleman 1989; Wagner et al.
2008) that display some territoriality and few interactions
between social units where intruders (i.e. individuals that are not
part of their community) are repelled aggressively (Mosser &
Packer 2009). Our results do not suggest that aggression is
common, although we observed agonistic displays (Martin 2007)
and bite wounds on some males instead of females (Porcher 2005),
which were not mating scars. However, it remains difficult to
distinguish between intra- and interspecific aggression. Such
individual preferences for other individuals that are not directly
related to reproduction, foraging or defence could be of benefitby
reducing aggression among neighbours within a defined range
(Pomeroy et al. 2005) and increasing familiarity (Ward & Hart
2003).
External Factors: Food Resources Distribution
On the north coast of Moorea, grouping could be related to the
presence of recreational provisioning (i.e. shark feeding), which
would promote aggregation by attracting individuals into
a restricted area where food is supplied. As highlighted by studies
on stingray feeding (Corcoran 2006; Semeniuk & Rothley 2008),
a localized supply of food brought in by tourists can alter an
animal’s behaviour by inducing more regular interactions between
conspecifics and by increasing fidelity to the feeding site (Clua
et al. 2010) as well as increasing spatial overlap (Atwood &
Weeks 2003) of solitary animals. If the observed spatial struc-
ture were simply an artefact of passive aggregations (or ‘loose
aggregations’) at patchily distributed resources, we would expect
that association strength would be higher, with temporally
unstable associations, at artificial feeding sites. We found that
density (i.e. community size) was higher for communities living
mainly within a feeding area (i.e. communities A, C and D with 54,
30 and 35 members, respectively) compared to community B,
which used a nonfeeding site (i.e. with 14 members; Fig. 3a). Mean
observed group size was also higher at feeding sites (i.e. 14.93 for
Site 2, 8.97 for Site 5 and 13.61 for Site 6) than adjacent sites
without feeding activities (i.e. 7.33 for Site 1, 9.30 for Site 4 and
4.60 for Site 7). However, site fidelity at these areas was not much
higher than in pristine environments for this species (Stevens
1984; Papastamatiou et al. 2009) and sharks that did not visit
feeding sites regularly also showed a high level of association
(HWI) but within smaller groups (Table 1). At provisioning sites,
shark communities showed nonrandom and temporally stable
associations. Provisioning may promote sociality by attracting
more potential social partners with increased food sources.
Community A from the main feeding site was the largest group
and showed a lower social cohesion. A high encounter rate created
by attraction to provisioned food may facilitate familiarity to
a large number of individuals thus allowing them to change
associates more regularly. Increased density can favour larger
group sizes, meaning that sociality may be density dependent in
this species, as shown in other animals (Macdonald 1983). As
blacktip reef sharks are top predators, the distribution of resources
such as food may influence their spatial distribution, but the
exclusivity of artificial feeding sites by a unique community
reflects the importance of sociality in mapping a home range, with
potential competition by exclusion between social groups (Lusseau
et al. 2006). Further research is needed to investigate the effect of
provisioning on sharks’behaviour and population structure, for
example by considering a direct comparison of provisioned and
nonprovisioned sites as implemented in other specific studies on
this topic (Semeniuk & Rothley 2008; Malkjovi
c & Côté 2011).
Social Variability
Blacktip reef sharks showed some variability in their sociality.
Individual sharks differed significantly in their level of gregarious-
ness and communities differed in their mean sociality. Community
A appeared to be the most diversely connected in the network (high
strength, eigenvector centrality, reach and affinity combined with
relatively low clustering) and had the highest number of members
(N¼54) subdivided into two subcommunities. Conversely,
community D showed a tighter clustering. These differences may be
related to individual characteristics, individual foraging or mating
strategies to maximize their fitness. Such variability within and
between communities highlights an interesting avenue for further
research in explanatory factors affecting social affiliation and
structure within this population. However, contrary to a previous
investigation into the role of sex and sex ratio in the social variability
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401398
of sharks (Jacoby et al. 2010), comparisons between male and female
centrality measures revealed no significant difference. Within
communities, the sex ratio was dominated by males or was equal as
in communities B and C, but this characteristic did not appear to
influence the social variability even if sharks tended toassort by sex.
Complex association patterns occur in numerous animal societies in
which cooperation has evolved (Krause & Ruxton 2002). Blacktip
reef sharks are known to form milling groups in an uncoordinated
fashion (see Appendix Fig. A1), which have generally been attrib-
uted to mutual attraction to a common resource such as food or
a refuge from predators (Motta & Wilga 2001). Evidence of coop-
erative feeding has been noted in several studies, and specifically by
Eibl-Eibesfeldt & Hass (1959) in blacktip reef sharks. Several indi-
viduals would herd a school of small fishes towards the shore,
providing food for all. During the study such cooperative hunting
was observed on two occasions, when a group of about four or five
blacktip reef sharks herded a school of fishes around a coral struc-
ture. However, it remains unknown whether sharks commonly
cooperate. The reciprocal altruism theory (Ohtsuki et al. 2006)
predicts that individuals would cooperate in small, tight groups (low
strength, high clustering) mostly composed of unrelated individ-
uals. Conversely, broader connectivity might exist if group members
are primarily kin, particularly in the context of cooperation (Sih et al.
2009). At the individual level, we can expect that individuals that
have higher mean relatedness to others in the group would also be
well connected in general (high strength, Sih et al. 2009). Although
interactions may occur between related individual sharks at the
juvenile stage within their nursery where they were born and grew
together (Guttridge et al. 2011), such assortment remains unknown
within adult populations. Further research should investigate the
relative importance of genetic relationships between individual
sharks as a potential factor shaping the structure of the network and
influencing social variability.
Conclusion
This study revealed for the first time that adults of a reef-
associated shark species formed communities that were main-
tained by nonrandom associations between specific individuals
with the capacity to form stable long-term social bonds. These
communities had different ranging patterns with little overlap
between them, individuals tending to assort even weakly by sex
and size. However, spatial assortment between individuals was
globally weaker than random expectations, suggesting that asso-
ciations resulted from social affiliation and were not an artefact of
either the sampling design or the spatial distribution of individuals.
These findings suggest that marine top predators such as blacktip
reef sharks display active preferences for specific individuals,
reinforcing current suggestions that familiarity may confer
substantial benefits in social marine fish. The grouping patterns
displayed by this shark species indicate that the structure of this
population does not reflect passive aggregations at specific
resources but rather developed from an active choice of individuals
similar in some ways to some other social animals. The decision to
form long-term social ties, developed through repeated interac-
tions, is therefore likely to have some ecological significance among
apex marine predators occupying a highly variable spatiotemporal
environment. Individual preferences and adaptation to local
conditions as well as demographic, ecological and anthropogenic
factors may explain the social variability between communities,
although it is possible that other factors, such as relatedness and
mating strategies, may shape the observed structure. Because
sharks are long-lived predators, the behavioural decisions they
make potentially impact their behaviour for periods of years to
decades.
Acknowledgments
This study was implemented thanks to the financial support of the
Direction à l’Environnement (DIREN) of French Polynesia, the Coor-
dination Unit of the Coral Reef Initiatives for the Pacific(CRISP
Program) and Proscience in French Polynesia. We are grateful to the
Centre de Recherche Insulaire et Observatoire de l’Environnement
(CRIOBE) staff fortheir technical support, as wellas the volunteers and
students who assisted in field data collection. We thank Hal White-
head, David Lusseau and Joanna Wiszniewski for all their support
with SOCPROG use and network analysis. We acknowledge Kathryn
Furby for manuscript reviewing and English language revisions.
References
Allee, W. C. & Dickinson, J. C. 1954. Dominance and subordination in the smooth
dogfish Mustelis canis (Mitchill). Physiological Zoology,27,356e364.
Atwood, T. C. & Weeks, H. P., Jr. 2003. Spatial home-range overlap and temporal
interaction in eastern coyotes: the influences of pair types and fragmentation.
Canadian Journal of Zoology,81, 1589e1597.
Baum, J. K., Myers, R. A., Kehler, D. G., Worm, B., Harley, S. J. & Doherty, P. A.
2003. Collapse and conservation of shark populations in the Northwest Atlantic.
Science,299, 389e392.
Bejder, L., Fletcher, D. & Brager, S. 1998. A method for testing association patterns
of social mammals. Animal Behaviour,56,719e725.
Bekoff, M., Daniels, T. J. & Gittleman, J. L. 1984. Life history patterns and the
comparative social ecology of carnivores. Annual Review of Ecology, Evolution,
and Systematics,15,191e232.
Borgatti, S. P., Everett, M. G. & Freeman, L. C. 2002. UCINET for Windows: Software
for Social Network Analysis. Harvard, Massachusetts: Analytic Technologies.
Buray, N., Mourier, J., Planes, S. & Clua, E. 2009. Underwater photo-identification
of sicklefin lemon shark, Negaprion acutidens, at Moorea (French Polynesia).
Cybium,33,21e27.
Cairns, S. J. & Schwager, S. J. 1987. A comparison of association indices. Animal
Behaviour,35, 1454e1469.
Carter, A. J., Macdonald, S. L., Thomson, V. A. & Goldizen, A. W. 2009. Structured
association patterns and their energetic benefits in female eastern grey
kangaroos, Macropus giganteus.Animal Behaviour,77, 839e846.
Chaverri, G., Gamba-Rios, M. & Kunz, T. H. 2007. Range overlap and association
patterns in the tent-making bat Artibeus watsoni.Animal Behaviour,73,
157e164.
Clark, E. 1959. Instrumental conditioning of lemon sharks. Science,130,217e218.
Clarke, K. R. & Warwick, R. M. 2001. Changes in Marine Communities: an Approach
to Statistical Analysis and Interpretation. 2nd edn. Plymouth: PRIMER-E.
Clua, E., Buray, N., Legendre, P., Mourier, J. & Planes, S. 2010. Behavioural response
of sicklefin lemon sharks Negaprion acutidens to underwater feeding for
ecotourism purposes. Marine Ecology Progress Series,414,257e266.
Clutton-Brock, T. H. 1989. Mammalian mating systems. Proceedings of the Royal
Society B,236,339e372.
Compagno, L. V., Dando, M. & Fowler, S. 20 05. Sharks of the World. Princeton, New
Jersey: Princeton University Press.
Corcoran, M. 2006. The effects of supplemental feeding on the activity space and
movement patterns of the southern stingray, Dasyatis americana, at Grand
Cayman, Cayman Islands. MSc. thesis, Nova Southeastern University.
Croft, D. P., James, R., Ward, A. J. W., Botham, M. S., Mawdsley, D. & Krause, J.
2005. Assortative interactions and social networks in fish. Oecologia,143,
211e219.
Croft, D. P., James, R. & Krause, J. 2008. Exploring Animal Social Networks.
Princeton, New Jersey: Princeton University Press.
Croft, D. P., Krause, J., Darden, S. K., Ramnarine, I. W. & James, R. 2009. Behav-
ioural trait assortment in social networks: patterns and implications. Behavioral
Ecology and Sociobiology,63, 1495e1503.
Croft, D. P., Madden, J. R., Franks, D. W. & James, R. 2011. Hypothesis testing in
animal social networks. Trends in Ecology & Evolution,26, 502e507, doi:10.1016/
j.tree.2011.05.012.
Domeier, M. & Nasby-Lucas, N. 2007. Annual re-sightings of photographically
identified white sharks (Carcharodon carcharias) at an eastern Pacific aggrega-
tion site (Guadalupe Island, Mexico). Marine Biology,150,977e984.
Duchesne, P., Étienne, C. & Bernatchez, L. 2006. PERM: a computer program to
detect structuring factors in social units. Molecular Ecology Notes,6, 965e967.
Ebert, D. A. 1991. Observations on the predatory behavior of the sevengill shark
Notorynchus cepedianus.South African Journal of Marine Science,11, 455e465.
Economakis, A. E. & Lobel, P. S.1998. Aggregation behaviour of the grey reef shark
Carcharhinus amblyrhynchos, at the Johnston Atoll, Central Pacific Ocean.
Environmental Biology of Fishes,51,129e139.
Eibl-Eibesfeldt, I. & Hass, H. 1959. Erfahrungen mit Haien. Zeitschrift für Tierp-
sychologie,16,733e746.
Frère, C. H., Krützen, M., Mann, J., Connor, R., Bejder, L. & Sherwin, W. B. 2010.
Home range overlap, matrilineal, and biparental kinship drive female associa-
tions in East Shark Bay bottlenose dolphins. Animal Behaviour,80, 481e486.
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401 399
Gaspar, C., Chateau, O. & Galzin, R. 2008. Feeding site frequentation by the pink
whipray Himantura fai in Moorea (French Polynesia) as determined by acoustic
telemetry. Cybium,32,153e164.
Gittleman, J. L. 1989. Carnivore group living: comparative trends. In: Carnivore
Behavior, Ecology, and Evolution (Ed. by J. L. Gittleman), pp. 183e207. Ithaca:
Cornell University Press.
Guttridge, T. L., Gruber, S. H., Gledhill, K. S., Croft, D. P., Sims, D. W. & Krause, J.
2009a. Social preferences of juvenile lemon sharks Negaprion brevirostris.
Animal Behaviour,78, 543e548.
Guttridge, T. L., Myrberg, A. A., Porcer, I. F., Sims, D. W. & Krause, J. 2009b. The
role of learning in shark behaviour. Fish and Fisheries,10, 450e469.
Guttridge, T. L., Gruber, S. H., Krause, J. & Sims, D. W. 2010. Novel acoustic
technology for studying free-ranging shark social behaviour by recording
individuals’interactions. PLoS ONE,5, e9324.
Guttridge, T. L., Gruber, S. H., DiBattista, J. D., Feldheim, K. A., Croft, D. P.,
Krause, S. & Krause, J. 2011. Assortative interactions and leadership in a wild
population of sharks. Marine Ecology Progress Series,423, 235e245.
Heupel, M. R. & Simpfendorfer, C. A. 2005. Quantitative analysis of aggregation
behaviour in juvenile blacktip sharks. Marine Biology,147, 1239e12 49.
Holland, K. N., Meyer, C. G. & Dagorn, L. C. 2009. Inter-animal telemetry: results
from first deployment of acoustic ‘business card’tags. Endangered Species
Research,10, 287e293.
Holyoak, M., Casagrandi, R., Nathan, R., Revilla, E. & Spiegel, O. 2008. Trends and
missing parts in the study of movement ecology. Proceedings of the National
Academy of Sciences, U.S.A.,105, 19060e19065.
Jacoby, D. M. P., Busawon, D. S. & Sims, D. W. 2010. Sex and social networking: the
influence of male presence on social structure of female shark groups. Behav-
ioral Ecology,21,808e818.
Jacoby, D. M. P., Croft, D. P. & Sims, D. W. 2011. Social behaviour in sharks and rays:
analysis, patterns and implications for conservation. Fish and Fisheries,
doi:10.1111/j.1467-2979.2011.00436.x, Published online 15 August 2011.
Jarman, P. J. 1974. The social organisation of the antelope in relation to their
ecology. Behaviour,48,215e267.
Klimley, A. P. & Nelson, D. R. 1981. Schooling of the scalloped hammerhead shark,
Sphyrna lewini, in the Gulf of California. Fisheries Bulletin,79, 356e360.
Krause, J. 1994. The influence of food competition and predation risk on size-
assortative shoaling in juvenile chub Leuciscus cephalus.Ethology,96,105e116 .
Krause, J. & Ruxton, G. D. 2002. Living in Groups. Oxford: Oxford University Press.
Krause, J., Croft, D. P. & James, R. 2007. Social network theory in the behavioural
sciences: potential applications. Behavioral Ecology and Sociobiology,62,15e27.
Krause, J., Lusseau, D. & James, R. 2009. Introduction to animal social networks.
Behavioral Ecology and Sociobiology,63, 967e973.
Krause, J., Wilson, A. D. M. & Croft, D. P. 2011. New technology facilitates the study
of social networks. Trends in Ecology & Evolution,26,5e6.
Lusseau, D., Wilson, B., Grellier, K., Hammond, P. S., Durban, J. W., Parsons, K. M.,
Barton, T. M. & Thompson, P. M. 2006. Quantifying the influence of sociality on
population structure in bottlenose dolphins. Journal of Animal Ecology,75,
14e24.
Lusseau, D., Whitehead, H. & Gero, S. 2008. Incorporating uncertainty into the
study of animal social networks. Animal Behaviour,75, 1809e18 15.
Macdonald, D. W. 1983. The ecology of carnivore social behaviour. Nature,301,
379e384.
McKibben, J. N. & Nelson, D. R. 1986. Patterns of movement and grouping of gray
reef sharks, Carcharhinus amblyrhynchos, at Enewetak, Marshall Islands. Bulletin
of Marine Science,38,89e110 .
Malkjovi
c, A. & Côté, I. M. 2011. Effects of tourism-related provisioning on the
trophic signatures and movement patterns of an apex predator, the Caribbean
reef shark. Biological Conservation,144 , 859e865.
Manly, B. F. J. 199 7. Randomization, Bootstrap, and Monte Carlo Methods in Biology.
New York: Chapman & Hall.
Martin, R. A. 2007. A review of shark agonistic displays: comparison of display
features and implications for sharkehuman interactions. Marine and Freshwater
Behavior and Physiology,40,3e34.
Meekan, M. G., Bradshaw, C. J. A., Press, M., McLean, C., Richards, A.,
Quasnichka, S. & Taylor, J. G. 2006. Population size and structure of whale
sharks Rhincodon typus at Ningaloo Reef, Western Australia. Marine Ecology
Progress Series,319,275e285.
Mosser, A. & Packer, C. 2009. Group territorialityand the benefits of sociality in the
African lion, Panthera leo.Animal Behaviour,78, 359e370.
Motta, P. J. & Wilga, C. D. 2001. Advances in the study of feeding mechanisms,
mechanics, and behaviors of sharks. Environmental Biology of Fishes,20,
131e156.
Myrberg, A. A. & Gruber, S. H. 1974. The behaviour of the bonnethead shark,
Sphyrna tiburo.Copeia,1974,358e373.
Nelson, D. R. & Johnson, R. H. 1980. Behavior of the reef sharks of Ranguiroa,
French Polynesia. National Geographic Society Research Reports,12,479e499.
Newman, M. E. J. 2006. Modularity and community structure in networks.
Proceedings of the National Academy of Sciences, U.S.A.,103,8577e8582.
Northcutt, R. G. 1977. Elasmobranch central nervous system organization and its
possible evolutionary significance. American Zoologist,17,411e429.
Ohtsuki, H., Hauert, C., Lieberman, E. & Nowak, M. A. 2006. A simple rule for the
evolution of cooperation on graphs and social networks. Nature,441, 502e505.
Papastamatiou, Y., Lowe, C. G., Caselle, J. E. & Friedlander, A. M. 2009. Scale-
dependent effects of habitat on movements and path structure of reef sharks at
a predator-dominated atoll. Ecology,90, 996e1008.
Papastamatiou, Y. P., Friedlander, A. M., Caselle, J. E. & Lowe, C. G. 2010. Long-
term movement patterns and trophic ecology of blacktip reef sharks (Carch-
arhinus melanopterus) at Palmyra Atoll. Journal of Experimental Marine Biology
and Ecology,386,94e102.
Pomeroy, P. P., Redman, P. R., Ruddell, S. J. S., Duck, C. D. & Twiss, S. D. 2005.
Breeding site choice fails to explain interannual associations of female grey
seals. Behavioral Ecology and Sociobiology,57, 546e556.
Porcher, I. F. 2005. On the gestation period of the blackfin reef shark, Carcharhinus
melanopterus, in waters off Moorea, French Polynesia. Marine Biology,146,
1207e1211.
Powter, D. M. & Gladstone, W. 2009. Habitat-mediated use of space by juvenile
and mating adult Port Jackson sharks, Heterodontus portusjacksoni, in Eastern
Australia. Pacific Science,63,1e14.
Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution,43, 223e225.
Sandell, M. 1989. The mating tactics and spacing patterns of solitary carnivores. In:
Carnivore Behavior, Ecology, and Evolution (Ed. by J. L. Gittleman), pp. 164e182.
Ithaca: Cornell University Press.
Semeniuk, C. A. D. & Rothley, K. D. 2008. Costs of group-living for a normally
solitary forager: effects of provisioning tourism on southern stingrays Dasyatis
americana.Marine Ecology Progress Series,357,271e282.
Sih, A., Hanser, S. F. & McHugh, K. 2009. Social network theory: new insights and
issues for behavioralecologists. BehavioralEcology and Social Biology,63,975e988.
Sims, D. W. 2003. Tractable models for testing theories about natural strategies:
foraging behaviour and habitat selection of free-ranging sharks. Journal of Fish
Biology,63,53e73.
Sims, D. W., Southall, E. J., Quayle, V. A. & Fox, A. M. 2000. Annual social behaviour
of basking sharks associated with coastal front areas. Proceedings of the Royal
Society B,267, 1897e1904.
Speed, C. W., Meekan, M. G., Field, I. C., McMahon, C. R., Stevens, J. D.,
McGregor, F., Berger, Y. & Bradshaw, C. J. A. 2011. Spatial and temporal
movement patterns of a multi-species coastal reef shark aggregation. Marine
Ecology Progress Series,429,261e275.
Springer, S. 1967. Social organization of shark populations. In: Sharks, Skates and
Rays (Ed. by P. W. Gilbert, R. F. Mathewson & D. P. Rall), pp. 149e174. Baltimore:
Johns Hopkins University Press.
Standora, E. A. & Nelson, D. R. 1977. A telemetric study of the behavior of free
swimming Pacific angel sharks Squatina californica.Bulletin of South California
Academy of Science,76,193e201.
Stevens, J. D. 1984. Life-history and ecology of sharks at Aldabra Atoll, Indian Ocean.
Proceedings of the Royal Society B,222,79e106.
Stevens, J. D., Bonfil, R., Dulvy, N. K. & Walker, P. A. 2000. The effects of fishing on
sharks, rays and chimaeras (chondrichthyans), and the implications for marine
ecosystems. ICES Journal of Marine Science,57, 476e494.
Striedter, G. F. 2005. Principles of Brain Evolution. Sunderland, Massachusetts:
Sinauer Associates.
Sundaresan, S. R., Fischhoff, I. R., Dunshoff, J. & Rubenstein, D. I. 2007. Network
metrics reveal differences in social organization between two fission-fusion
species, Grevy’s zebra and onager. Behavioral Ecology,151,140e149.
Wagner, A. P., Frank, L. G. & Creel, S. 2008. Spatial grouping in behaviourally
solitary striped hyaena, Hyaena hyaena.Animal Behaviour,75,1131e1142.
Ward, A. J. W. & Hart, P. J. B. 2003. The effects of kin and familiarity on interactions
between fish. Fish and Fisheries,4, 348e358.
Ward-Paige, C., Mills Flemming, J. & Lotze, H. K. 2010. Overestimating fish counts
by non instantaneous visual censuses: consequences for population and
community descriptions. PLoS ONE,5, e11722.
Wearmouth, V. J. & Sims, D. W. 2008. Sexual segregation in marine fish, reptiles,
birds and mammals: behaviour patterns, mechanisms and conservation
implications. Advances in Marine Biology,54,107e170.
Wetherbee, B. M. & Cortes, E. 2004. Food consumption and feeding habits. In:
Biology of Sharks and their Relatives (Ed. by J. C. Carrier, J. A. Musick &
M. R. Heithaus), pp. 223e244. Boca Raton: CRC Press.
Whitehead, H. 1995. Investigating structure and temporal scale in social organi-
zations using identified individuals. Behavioral Ecology,6,199e208.
Whitehead, H. P. 1999. Testing association patterns of social animals. Animal
Behaviour,57, F26eF29.
Whitehead, H. & Dufault, S. 1999. Techniques for analyzing vertebrate social
structure using identified individuals: review and recommendations. Advances
in the Study of Behavior,28,33e74.
Whitehead, H. 2008. Analysing Animal Societies: Quantitative Methods for Vertebrate
Social Analysis. Chicago: University of Chicago Press.
Whitehead, H. 2009. SOCPROG programs: analyzing animal social structures.
Behavioral Ecology and Sociobiology,63,7
65e778.
Whitehead, H., Bejder, L. & Ottensmeyer, C. A. 2005. Testing association patterns:
issues arising and extensions. Animal Behaviour,69,e1ee6.
Wiszniewski, J., Allen, S. J. & Moller, L. M. 2009. Social cohesion in a hierarchically
structured embayment population of Indo-Pacific bottlenose dolphins. Animal
Behaviour,77,1449e1457.
Wiszniewski, J., Lusseau, D. & Möller, L. M. 2010. Female bisexual kinship ties
maintain social cohesion of a dolphin network. Animal Behaviour,80, 895e904.
Wolf, J. B. W., Mawdsley, D., Trillmich, F. & James, R. 2007. Social structure in
a colonial mammal: unravelling hidden structural layers and their foundations
by network analysis. Animal Behaviour,74, 1293e1302.
Yopak, K. E., Lisney, T. J., Collins, S. P. & Montgomery, J. C. 2007. Variation in brain
organization and cerebellar foliation in chondrichthyans: sharks and holloce-
phalans. Brain Behavior and Evolution,69, 280e300.
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401400
Appendix
Figure A1. Blacktip reef shark social behaviours observed during the study: (a) two mature males following; (b) two mature males paralleling; (c) four adult sharks milling.
Following swimming: two or more sharks swimming nose to tail within four body lengths of each other; one mimics the directional changes of the leader. Parallel swimming: two
or more sharks observed swimming in parallel within about two body lengths of each other and exhibiting the same directional changes in swimming behaviour. Milling group or
loose aggregation: two or more sharks swimming together but not exhibiting any coordinated directional changes. Photos: Johann Mourier.
J. Mourier et al. / Animal Behaviour 83 (2012) 389e401 401