Patterns of Aggression Among Captive American
Flamingos (Phoenicopterus ruber)
Mitchell G. Hinton,1Annie Bendelow,1Samantha Lantz,1Tina W. Wey,2Lee Schoen,3Robin Brockett,3
and Jordan Karubian1*
1Department of Ecology and Evolutionary Biology, Tulane University, New Orleans, Louisiana
2Department of Environmental Science and Policy, University of California at Davis, Davis, California
3Audubon Zoo, Audubon Nature Institute, New Orleans, Louisiana
Many species of flamingo are endangered in the wild but common in zoos, where successful captive breeding programs are
a management priority. Unlike their counterparts in the wild, captive flamingo individuals are easy to mark and follow,
facilitating longitudinal data collection on social dynamics that may affect reproduction. We studied a captive group of
American Flamingos at the Audubon Zoo in New Orleans, LA to document patterns of aggression between individuals
during the onset of breeding. We used a social network approach to test whether overall aggression would be higher during
courtship or following establishment of pair bonds. Aggression was higher following pair bond establishment than during
courtship, suggesting that individuals in our study population may compete more intensely for resources such as nesting
sites than for mates. We also found that males were more aggressive than females during all stages of the study period and
that there was a positive relationship between age and aggression in males during the pair‐bond stage. We discuss these
findings in light of management practices for captive populations of flamingos and general patterns of aggression in social
animals. Zoo Biol. 32:445–453, 2013.
© 2013 Wiley Periodicals Inc.
Keywords: bird; captive animal; social network theory; UCINET; zoo management
(e.g., Cezilly et al., 1995; Bechet and Johnson, 2008), and
four of six recognized flamingo species are in a state of
decline in the wild (Ottenwalder et al., 1990; IUCN, 2011).
For this reason, ex situ conservation methods such as captive
breeding programs designed to maintain long‐term retention
of genetic diversity may play an important role in the
Unfortunately, flamingos have been historically difficult to
breed in captivity (Farrell et al., 2000). As such, an improved
understanding of flamingo reproductive behavior is likely
to improve ex situ management of this species (Cezilly
et al., 1996). Managers often use information on behavior
of individuals to inform captive management practices
(Wielebnowski, 1998), but wild flocks of flamingos often
and studying individuals within these groups (Allen, 1959;
Bildstein et al., 1993; King, 2000). Because so little is known
about these birds in the wild, it is useful to gather detailed
data on the reproductive behavior of captive individuals.
Captive flamingos are easy to observe and many potentially
confounding variables (e.g., diet and access to carotenoids)
can be controlled (King, 2000), leading to informed
behavioral studies that may also increase our understanding
of wild populations.
Intraspecific aggression among group living animals
in captive environments is often associated with stress
and reduced reproductive output (Cezilly et al., 1996;
Conflicts of interest: None.
*Correspondence to: Jordan Karubian, Department of Ecology and
Evolutionary Biology, Tulane University, 400 Boggs Building, New
Orleans, Louisiana 70118. E‐mail: firstname.lastname@example.org
Received 29 November 2012; Revised 25 April 2013; Accepted 06 May
Published online 11 June 2013 in Wiley Online Library
© 2013 Wiley Periodicals, Inc.
Zoo Biology 32: 445–453 (2013)
Sandell, 1998; Ramos, 2003; Jakubas, 2004; Ellis and
Good, 2006; Gonzalez‐Voyer et al., 2007). Such aggressive
behavior is found in all species of both captive and wild
populations of flamingos and is likely associated with large
group sizes (Bildstein et al., 1993; Farrell et al., 2000; Perdue
et al., 2011). In flamingos, aggression has been found to
et al., 2000) and to negatively affect the foraging success
of juveniles in the wild (Cezilly et al., 1996). However, many
questions remain about how aggressive behavior is exhibited
leading up to and during breeding. In particular, aggression
during the onset of breeding may arise from competition over
mates, or competition over limiting resources such as nest
sites or food. Distinguishing between these two alternatives,
as well as identifying any age‐ or sex‐related propensities to
engage in aggressive interactions, can help curators to more
successfully manage captive populations.
Flamingo breeding has been described as opportu-
nistic (Studer‐Thiersch, 2000). Group courtship displays
by reproductively active flamingos are induced by
favorable environmental conditions (e.g., high nutrient
availability and warm temperatures) and onset of repro-
ductive condition (Cezilly et al., 1995, 1996; Farrell
et al., 2000). Shortly following these displays, individuals
then form pair‐bonds for the purposes of nest building,
chick rearing, and protection from intraspecific aggressors
(Stevens et al., 1992; Shannon, 2000; Perdue et al., 2011).
Pair‐bonds do not typically persist for more than one
season (Cezilly et al., 1997), meaning that courtship
display and pair establishment are likely to occur every
breeding season. For this reason, patterns of aggression
during these periods are likely to have important
consequences for breeding success.
Four peer‐reviewed studies have been published on the
aggressive behavior of captive flamingos; three on Chilean
Flamingos (Phoenicopterus chilensis), and one on American
Flamingos (P. ruber). Farrell et al.(2000) observedincreased
differences were observed in males versus females in either
period. In a second study conducted only during the breeding
season, Chilean Flamingo adults were more aggressive than
juveniles, and bonded individuals (in small groups of two or
more) were significantly more aggressive than solitary birds
(Perdue et al., 2011). Stevens et al. (1992) suggested that egg
losses in Chilean Flamingos were a result of intraspecific
aggressive interactions at nesting mounds. Finally, Anderson
et al. (2009) determined that preferred neck‐resting position
predicts aggression in American Flamingos. Beyond these
studies, aggressive conspecific interactions among captive
flamingos and our understanding of how social interactions
shift as individuals transition from nonbreeding to breeding
status have gone largely unexplored.
In the present study, we characterize patterns of
aggression among captive American flamingos with a focus
on how these patterns change as animals transition from
season progresses (Farrell
courtship to pair establishment. To do so we used Social
Network Analysis (SNA), which provides useful tools for
gregariousspecies (Croft et al.,2008; Wey et al., 2008). SNA
is based on modeling a system of nodes (that represent
individuals) and the ties (interactions or relationships) that
connect those nodes. SNA introduces the possibility of
quantifying individual centrality (importance or position)
individuals. Wey et al. (2008) suggests that SNA may
improve our understanding of social structure by examining
measures of social interaction directly (as opposed to
traditional measures of group size, mating system, and rates
of interaction). This allows for both qualitative and
quantitative analyses. As opposed to general comparisons
of aggression rates, SNA offers the advantage of tracking
interactions among specific individuals with known charac-
teristics, providing a more complete understanding of social
variation among individuals within the group. For example it
can provide insight into patterns in which individuals direct
their aggression (e.g., whether males direct aggression more
often to other males, females, or randomly).
To date, researchers have utilized network theory to
describe social structure in a diverse array of taxa in the wild
including mammals, fish, and birds (reviewed in Croft et al.,
to captive animal management. McCowan et al. (2008), for
instance, used SNA to study aggression in captive rhesus
macaques (Macaca mulatta). Using event‐sampling method-
ology, they were able to examine metrics for groom
reciprocity, subordination, displacement, and aggression.
They concluded that aggression and subordination are
affected by group composition (i.e., age and sex) and that
manipulation might be a useful tool in reducing aggression
levels within these captive groups. In a separate study, White
between social complexity and reproductive success in
captive cowbirds (Molothrus ater). They compared individ-
uals conditioned in dynamic social groups in which
composition was consistently altered, and stable social
groups in which composition remained fixed. Upon mixing
dynamic‐conditioned and stable‐conditioned males, they
found that dynamic‐conditioned individuals showed greater
reproductive success. To date, few studies have utilized SNA
to map avian aggressive social structure, but as the above
studies evidence SNA can be a powerful tool in behavioral
research as well as captive animal management.
In our study, we mapped and analyzed social
interactions between individual American Flamingos at the
Audubon Zoo in New Orleans, LA. Our first goal was to
of importance for managers because it may provide insight
into whether animals are competing more intensely for mates
(i.e., during courtship), or resources once pair formation has
Hinton et al.
occurred (i.e., for nesting sites or food). Our second goal was
to characterize patterns of aggression in relation to sex and
age, a question of importance for managers. Based on
findings reported for Chilean flamingos, we predicted that
there should be little difference in aggressive behavior
between males and females and that age should be associated
with increased aggression in both sexes.
We studied a group of 34 (16 males and 18 females)
individually marked American Flamingos from February 20
has bred successfully in past years and exhibited courtship
data). We constructed a standardized ethogram based on
preliminary observations (5hr) and recorded individual
agonistic encounters using a combination of focal‐animal
focal animal samples, we recorded consecutive 5‐min
sampling periods on randomly selected individuals between
a total of 17.17hr of observation throughout our study. We
selected individuals for focal observation using a random
number generating procedure, and at least one focal
observation was conducted for each individual within the
population (5.32?2.81SD focal observations per individual;
range¼1–11). During focal‐animal observations, we also
made opportunistic observations of aggressive interac-
tions by non‐focal animals. We did not start recording
opportunistic observations of aggression until March 11
causing our traditional (non‐SNA) t‐test analysis of
overall rate of aggression to incorporate data between
March 11 (as opposed to February 23), and April 12;
aggression was relatively rare prior to March 11. We
noted the initiators and recipients of all agonistic actions
including directed pecking/squawking and stand‐offs (in
which two or more individuals mutually exchanged
extended vocalizations and waved their heads in an
aggressive display; Studer‐Thiersch, 2000). A majority
(58%) of our observations of aggressive interactions were
opportunistic (i.e., did not involve a focal individual). As
such, we are confident that differences in number of
observations among individuals did not heavily impact
our ability to sufficiently capture levels of aggression for
We created separate networks for the courtship period
(February 20–March 22 2012; Fig. 1a) and the pair‐bond
period (March 23–April 12; Fig. 1b). As such, our data
collection period during the courtship stage was slightly
longer than that during the pair‐bonding stage (29 days and
9.5hr of observation vs. 22 days and 7.4hr of observation,
respectively). The cut‐off date between courtship and pair‐
bond was determined by visually assessing the proportion of
pair‐bonded individuals and the relative frequency of
courtship displays (Fig. 2). Our cut‐off date of March 22
corresponds to the time at which courtship displays had
halted, a majority of individuals were pair‐bonded, and
zookeepers created artificial mounds within the exhibit to
stimulate nest building. Modest shifts (i.e., ?7 days) in the
cut‐off date between these two stages did not change our
We mapped directed, weighted networks (i.e., inter-
actions had an initiator and recipient, and ties between
individuals were scored based on the number of interactions
betweenthoseindividuals). Eachnode(point onthenetwork)
represents a specific individual and each tie connecting the
nodes represents a directed aggressive interaction. Using
UCINET software we were able to quantify and compare
aggressive behavior between groups of individuals and
between the temporally disparate networks (Borgatti
et al., 2002; Hanneman and Riddle, 2005). Creating a
at Audubon Zoo, based on agonistic interactions. Nodes indicate
individuals and ties indicate directed aggressive interactions (ties
include arrows signifying the direction of the interaction). The
thickness of ties indicate weight of interaction (thinnest lines¼1
interaction; thickest lines¼4 interactions). Panel (a) depicts the
courtship period network, during which juveniles and females were
involved in relatively few agonistic initiations, and a few individual
males exhibited relatively large amounts of aggression and were
therefore central within the network. Panel (b) depicts the pair‐bond
period network, which was much denser than the courtship period,
reflecting higher levels of overall aggression. As in the courtship
network, however, juveniles and females remained closer to the
outskirts andmales tendedtobe morecentraldue togreater amounts
of aggressive initiations. See text for more details.
Graphical depictions of American flamingo social network
Aggression in Captive American Flamingos
directed social network allowed us to analyze which
flamingos were initiating and receiving aggression. In
addition, we also used a more traditional analysis (we
analyzed rates of aggressive interactions defined as a certain
number of aggressive interactions observed within an
established period of time) to determine the overall trend
achieved using network analysis. Our traditional measure of
5min) unlike SNA, did not allow for comparisons using
extra‐focal observations so we only included the aggressive
interactions of our focal individuals (this did not lead to
different results for overall group levels of aggression).
We quantified aggression using standard network
metrics at the individual level (network terminology defined
in Table 1). These measures were chosen to capture different
aspects of individual social connectedness and network
structure. We also calculated network‐level versions of
individual‐level measures (Table 1). One key measure,
network degree centralization, provides an index of variance
among individuals in the number of ties (the higher the
variance, the less equally aggressive interactions are
distributed among individuals). Network data generally
violates assumptions about independence of data points, so
significance is more accurately determined by generating a
distribution from the data, rather than using standard
significance tests (Croft et al., 2011). For this reason,
UCINET evaluates significance via bootstrapping permuta-
tion calculations without test statistics, or degrees of freedom
(Hanneman and Riddle, 2005). We also determined differ-
ences in aggression between males and females, and between
the pre‐breeding network and breeding network using
permutation t‐tests. We square‐root transformed our meas-
ures of degree centrality (out‐degree and in‐degree; Table 1)
the likelihood that aggressive interactions were not random
with respect to sex (whether males or females directed
aggression to a particular sex), we utilized an E–I Index
(external–internal index) permutation test with males and
females as our classes.
We constructed two social networks based on aggres-
sive interactions, one for the courtship period and one for the
pair‐bond period (Fig. 1). We observed a significant increase
in overall aggressive interactions in the (earlier) courtship
network versus the (later) pair‐bond network (one‐tailed t‐
test; t¼?4.78, df¼33, P<0.05), as measured by density.
The pair‐bond network also had a significantly higher
incidence of aggressive encounters initiated (as measured
by mean out‐degree; P<0.0001; mean total out‐degree?
SE; courtship: 2.25?0.15 vs. pair‐bond: 3.54?0.20), and
received (as measured by mean in‐degree; P<0.0001; mean
total in‐degree?SE; courtship: 2.33?0.10 vs. pair‐bond:
3.59?0.17). A more traditional analysis of differences in
captive American flamingos in the Audubon Zoo. The dotted line at March 22 shows the breakpoint we used to distinguish between the
courtship and pair‐bond networks analyzed. The star indicates the date at which dirt mounds were created by keepers in order to stimulate
nest building (March 16).
The proportion of pair‐bonds (right axis) increased, and frequency of courtship displays (left axis) decreased during our study of
Hinton et al.
TABLE 1. Social network metrics used to characterize patterns of aggression among captive American Flamingos
Individual level metrics
Number of ties originating from an
A metric for determining which
individuals initiate aggressiveinteractions
Males more aggressive than females
during pair‐bond and courtship
Number of ties directed toward an
A metric for determining which
individuals are receiving the most
Equal levels of aggression reception
among males and females duringcourtship and pair‐bond
Network level metrics
Number of realized ties divided by the
number of possible ties within thenetwork
Used to analyze the overall levels of
aggression within our population andcompare aggression levels temporally
Pair‐bond period exhibited more
aggression than courtship period
The tendency of ties to exist between
similar individuals (of the same sex in
the present study)
Used to ascertain which individuals are
prime targets for aggression by a
specified group of other individuals
Males and females distributed aggression
without regard to sex in both courtship
and pair‐bond periods
The density of ties among the neighbors
(those connected via ties) of the focal
A means of determining whether
individuals focused aggression on a
particular subset of other individuals,
without regard to sex of those
Pair‐bond period had greater
within‐group formation than
An index of variance in the number of ties
Provides measures of how evenly
aggressive interactions are initiatedand received
Relatively low measures throughout the
study indicate that aggression was
fairly equally distributed
Shown is the name and definition of each metric (following Wey et al., 2008, Hanneman and Riddle, 2005), as well as a short description of the way in which we applied the metric to answer
questions about aggression in flamingos, and a summary of results.
Aggression in Captive American Flamingos
rates of aggression between the two stages corroborates an
overall increase in aggression during the pair‐bond period
versus the courtship period (one‐tailed t‐test; df refer to
number of observation days: t¼?2.34, df¼11, P¼0.019).
The mean clustering co‐efficient (defined in Table 1) of the
pair‐bond network was significantly larger than that of the
courtship network (P¼0.0001; mean clustering co‐efficient
?SE; pair‐bond: 0.43?0.019 vs. courtship: 0.17?0.02),
congruent with a higher degree of subgroup formation
(formation of groups within the population) after pair‐bonds
Males were more aggressive than females in both
stages of the study. During the courtship period, males
initiated significantly more aggression (mean out‐degree?
SE; male 2.72?0.24 vs. female 1.84?0.10; P¼0.0008;
Fig. 3a). During the pair‐bond period, when overall
aggression was higher, males continued to exhibit higher
levels ofaggression (mean
4.41?0.21 vs. female 2.76?0.19; P<0.0001; Fig. 3b).
Centralization provides measures of how evenly
aggressive interactions are initiated and received among
individuals in a network, respectively; a higher value
indicates greater variance in the distribution of these events
across individuals. Network out‐degree and in‐degree
centralization provide insights into finer‐grained relation-
ships of how age may interact with sex to affect patterns of
aggression during these two time periods. Network out‐
degree centralization were similar between the courtship, and
pair‐bondperiods (16.58% vs. 7.87%, respectively), suggest-
ing an intermediate level of variance in the distribution of
aggression in both stages. However, the difference between
the two is consistent with our finding that during courtship a
few individuals initiated many aggressive interactions
whereas during the pair‐bond period aggressive behavior
was more evenly distributed (see Fig. 3).
During the courtship period, there was no relationship
between age and aggression in males (Fig. 4a; linear
regression; r2¼0.13; t¼1.46, df¼14, P¼0.17). Similarly,
females showed no relationship between age and aggression
(Fig. 4a; linear regression; r2¼0.12; t¼?1.28, df¼12,
P¼0.22). During the pair‐bond period, there was a
significant increase in aggression (out‐degree) with age
among males (Fig. 4b; two‐tailed t‐test; r2¼0.29; t¼2.42,
df¼14, P¼0.03), but again there was no relationship
test; r2¼0.01; t¼0.35, df¼12, P¼0.73). These findings
suggest that although there was no age‐based difference in
aggression among males and females during the courtship
period, as the season progressed a sub‐set of relatively old
males exhibited increased levels of aggression relative to the
females and the rest of the males.
In contrast, there was a more equitable distribution of
reception of aggressive interactions across the study period,
with no easily discernible outliers during thecourtship period
aggressive initiations (as measured by out‐degree; see Table 1).
a: During the courtship period there was a right skewed distribution
with male outliers displaying relatively large amounts of aggressive
initiations. b: Comparatively there was less skew during the pair‐
bond period. Refer to Table 1 for a more detailed description of out‐
degree and its applications within this study.
Histograms showing the numbers of individuals initiating
female American flamingos in the Audubon Zoo during (a) the
courtship period, during which there was no significant relationship
among males and females and (b) the pair‐bond period, during
which aggression was positively associated with age among males,
but not females.
Linear regression for out‐degree versus age in male and
Hinton et al.
versus pair‐bond period (in‐degree centralization 4.09% for
the courtship network vs. 8.66% for the pair‐bonding
network). This result is corroborated in both courtship and
pair‐bond periods by mean reception of aggressive interac-
tion, and homophily, the later of which we used in this study
to assess the propensity for outgoing ties to be directed to
members of the same sex. During the courtship stage, there
was no difference in reception (mean in‐degree?SE; male
directed aggression towards males and females uniformly
(mean homophily ?SE; males¼0.61?0.069, females
¼0.43?0.083; E–I Index, P¼0.118). This held true during
the pair‐bond period as there was, again, no difference in
reception (mean in‐degree?SE; male 3.87?0.21 vs. female
3.35?0.25; P¼0.42), and no significant deviation from
randomness in proportion of same‐sex directed outgoing ties
¼0.55?0.06; E–I Index, P¼0.878). Thus, aggression
was directed withoutregard tosexoftherecipient throughout
our study period.
In the captive population of American flamingos we
studied at the Audubon zoo, males were consistently more
aggressive than were females throughout our study period,
which included the early stages (i.e., courtship display and
pair formation) of the breeding season. This was counter to
our predictions, and differed from a previous study that
found equivalent levels of aggression in male and female
Chilean Flamingos (Perdue et al., 2011). Perdue et al.
(2011) began their study after pair‐bond formation,
whereas, our study concluded with pair‐bond formation.
It is possible that aggression levels between sexes begin to
reach a state of equilibrium after pair‐bonds are fully
formed and nest construction begins; nest protection and
chick rearing may become a primary concern for both
males and females alike. To our knowledge, no studies
have explored sex differences in aggression during
incubation and chick rearing. Perdue et al. (2011) also
found that pair‐bonded Chilean Flamingos were signifi-
cantly moreaggressive than solitary individuals.Pair‐bond
status, however, was fairly uniform within our study group
(32 of 34 flamingos were in a pair‐bond).
A second major finding of our study is the documenta-
tion of a temporal shift in social structure during the onset of
breeding as the study population transitioned from courtship
to established pair‐bonds. More specifically, we observed
significant increases in aggression over time as the onset of
the breeding season progressed. This was in agreement with
another study in which levels of aggression, during the
breeding season of captive Chilean Flamingos, increased as
time progressed (Farrell et al., 2000). Additionally, 13% of
eggs laid were destroyed due to aggressive interaction
between neighbors. Farrell et al. (2000) concluded that the
observed increases in aggression were most likely due to an
increased demand for nest/chick defense; they suggested that
artificial nest mounds at least 1.5 neck lengths apart be built
prior to egg deposition in order to alleviate aggression.
significant increases in agonistic interaction may too be a
We also documented a significant positive correlation
between age and aggression in males during the pair‐bond
stage. Though a previous study found differences in
et al., 1991), no other study has looked at patterns of
aggression in relation to exact age in flamingos. Our finding
suggests that age is associated with a males’ tendency to be
aggressive. Collis and Borgia (1992) found a relationship
between age and aggressive dominance in satin bowerbirds
(Ptilonorhynchus violaceus), despite androgen supplementa-
tion; they suggest that this is consistent with the idea that
experience plays a large role in aggression and dominance. It
ispossible thataccumulationofexperience inmale flamingos
influences willingness to invest energy in agonistic encoun-
ters. More longitudinal studies, however, could give further
insight into the potential relationship between age and
By assessing these trends in a SNA framework, we
were able to gain novel insights into more subtle social
approaches tobehavior may overlook. At theoverall network
level, though both the courtship network, and pair‐bond
network had a low clustering co‐efficient, the courtship
network had a significantly lower clustering co‐efficient than
the pair‐bond network. Because clustering co‐efficient is a
measure of the density of interactions between an individual
and its neighbors, our results indicate that there were fewer
This difference can most easily be explained by pair‐bond
formation following courtship. Based on qualitative obser-
vations, bonded individuals seemed to retaliate to aggression
directed toward their partners. We also observed pair‐bonded
individuals staking claim to artificial nesting mounds, which
seemed to lead to increased aggressive interactions with
spatially adjacent pairs. Regardless, as the lack of differenti-
ation in reception in both networks suggests, individuals still
distributed aggression relatively homogenously, and any
potential subgroups that may have formed were most likely
due to spatial arrangement, not social position. This, paired
with a seemingly indiscriminate distribution of aggressive
initiations rather than targeted initiations towards specific
individuals, suggests that elevated aggression may result
from competition for nesting resources rather than competi-
tion for mates.
The social network approach also yielded interesting
insights at the individual level. Measures of centrality at the
individual level (i.e., comparisons of individual measure-
ments) may not necessarily lend statistically pertinent
information (James et al., 2009), but we think it noteworthy
to highlight several aspects of this particular system. For
Aggression in Captive American Flamingos
most central individuals remained central throughout, sug-
gesting that social structure, in terms of individual positions
within the network, remained relatively stable as time
progressed. In the courtship network, the individual that
initiated the highest amounts of aggressive interactions we
recorded was a male whose closest competitors during the
aggressive flamingo displayed the second highest level of
aggression in the pair‐bond network, while the most
the most aggressive male during the courtship period
paired with another male during the pair‐bond period.
Also, during the pair‐bond network, a few females
displayed levels of aggression nearly equivalent to those
of the males in that network. Defending their pair bonded
mate may have contributed to this increase in female
The higher variance in aggressive initiations initiated
exhibit, demonstrates that certain individuals exhibit higher
levels of aggression. Furthermore, highly aggressive individ-
uals could heavily affect overall network structure (Sih and
Watters, 2005), though this may be less important once pair‐
bondsform inourstudygroup.Bildstein etal.(1991)showed
that feeding behavior is hindered by aggressive interaction,
particularly in younger individuals who received a greater
proportion of aggressive displays. Received aggression may
negatively affect the foraging activity of breeding individua-
ls, especially females who require a disproportionate amount
limited copulation success due to pinioning (King, 1994),
may contribute to reproductive failure (Farrell et al., 2000).
Forming a better understanding of the differences in
aggression levels and social structure between wild and
captive flocks might provide more insight into management
techniques that would maximize group health.
Spatial structure of the exhibits might impact aggres-
pressures in captive flamingos (Farrell et al., 2000). Our
particular exhibit was approximately 14m?15m with a
single feeding area, and housed 34 individuals, a density of
approximately one animal per 6.2m2(the density of
individuals in other studies was not reported). Though wild
populations of flamingo are typically dense (Tuite, 1979),
they are much larger in number than captive populations.
Estevez et al. (2003) show that aggression levels decrease
with increasing group size in domestic fowl; small captive
groups of flamingos may experience larger‐than‐natural rates
of aggression. But rather than manipulation of group sizes,
increases in exhibit size, or foraging space could ameliorate
aggression by providing more areas for refuge of subordinate
individuals, and potentially relieving competitive pressures
regarding nesting resources. However, this possibility
remains untested, and increasing exhibit size may not be
practical in every zoo. We suggest further research to
compare the social structure of varying groups inhabiting
disparately sized exhibits.
Future studies could also explore the effects of the
experimental removal of highly central (i.e., aggressive)
individuals on the social structure of the population. Due to
flamingos heightened response to disturbance during pre‐
breeding (courtship and pair establishment), and breeding
seasons, any such removal should occur during the non‐
impact breeding. Flack et al. (2006) show significant
alterations in primate social structure upon removal of highly
central individuals; however, it is difficult to determine
whether, the removal of an individual within our flamingo
population will elicit significant shifts in social structure in
accommodation for the missing individual (another may take
the vacant position). Before proceeding with removals, it
would be important to assess the reproductive output of these
central individuals and the colony as a whole when they are
present, in order to have a baseline for comparison. Because
individuals differ in their level and consistency of aggression
be an exciting future research direction with management
implications, especially in small captive populations with
detailed data on individuals. Using SNA, managers may be
able to identify highly aggressive or stressed individuals and
take appropriate management actions.
wild flocks, especially with regard to flock size and foraging
needs. Still, zoo studies remain a useful tool in understanding
animal behavior, and flamingos are a prime example of why
captive research is necessary and can yield valuable results
SNA metrics could prove important in revealing potential
display disproportionate amounts of aggression and which
individuals receive this aggression), as well as aiding in the
development of techniques that might serve to promote
population health (e.g., determining sufficiency of exhibits in
providing areas for refuge from aggression). The indiscrimi-
nate initiation of aggressive interaction along with the
increased level of aggression during the pair‐bond stage
suggests that elevated aggression may be linked to competi-
tion for nesting resources, and not competition for mates.
Because aggression might not be obligate for successful
breeding, efforts to reduce aggressive interactions in captive
populations could be useful in promoting individual health
and reproduction. Furthermore, keeping in mind that highly
central individuals appeared to remain central throughout the
onset of the breeding season, we suggest further network
non‐breeding season on individual reproductive success
during the breeding season. Regardless of the cause of
aggression, longitudinal studies may be useful in predicting
future mating success, as was predicted by SNA of social
structure in long‐tailed manakins (Chiroxiphia linearis)
Hinton et al.
(McDonald, 2007). Historically, zoos have had trouble with
successfully mating flamingos (Farrell et al., 2000); if able
to successfully predict or promote reproductive success,
network theory and more detailed analyses of social structure
for conservation if wild populations continue to decline.
on an earlier draft of this article. We would also like to thank
the Audubon Zoo and Audubon Nature Institute for allowing
us to conduct this research, students of Tulane University’s
Experimental Animal Behavior Course (EBIO369) for
helpful feedback throughout the various stages of this study,
and Dr. Scott Walter for his help with figure preparation.
Allen RP. 1959. The flamingos: their life history and survival. New York:
National Audubon Society.
Altmann J. 1974. Observational study of behavior: sampling methods.
Anderson MJ, Williams SA, Bono AJ. 2009. Preferred neck‐resting position
predicts aggression in Caribbean flamingos (Phoenicopterus ruber).
Bechet A, Johnson AR. 2008. Anthropogenic and environmental determinants
ty in the Camargue (Rhone delta, southern France). Ibis 150:69–79.
Bildstein KL, Frederick PC, Spalding MG. 1991. Feeding patterns and
aggressive behavior in juvenile and adult American flamingos. Condor
Bildstein KL, Golden CB, McCraith BJ, Bohmke BW, Seibels RE. 1993.
Feeding behavior, aggression, and the conservation biology of flamingos:
Borgatti SP, Everett MG, Freeman LC. 2002. Ucinet for windows: software
for social network analysis. Harvard, MA: Analytic Technologies.
Cezilly F, Boy V, Green RE, Hirons GJM, Johnson AR. 1995. Interannual
variation in greater flamingo breeding success in relation to water levels.
and breeding probability in greater flamingos. Ecology 77:1143–1150.
Cezilly F, Boy V, Tourenq C, Johnson AR. 1997. Age‐assortative pairing in
the Greater Flamingo Phoenicopterus ruber roseus. Ibis 139:331–336.
Collis K, Borgia G. 1992. Age‐related effects of testosterone, plumage, and
experience on aggression and social dominance in juvenile male satin
bowerbirds (Ptilonorhynchus violaceus). The Auk 109:422–434.
Croft DP, James R, Krause J. 2008. Exploring animal social networks.
Princeton, NJ: Princeton University Press.
Croft DP, Madden JR, Franks DW, James R. 2011. Hypothesis testing in
animal social networks. Trends Ecol Evol 26:502–507.
EllisJC,GoodTP.2006.Nestattributes, aggression,and breedingsuccessof
gulls in single and mixed species subcolonies. Condor 108:211–219.
Estevez I, Keeling LJ, Newberry RC. 2003. Decreasing aggression with
increasing group size in young domestic fowl. Appl Anim Behav Sci
Farrell MA, Barry E, Marples N. 2000. Breeding behavior in a flock of
Chilean flamingos (Phoenicopterus chilensis) at Dublin zoo. Zoo Biol
Flack JC, Girvan M, Waal FBM, Krakauer DC. 2006. Policing stabilizes
construction of social niches in primates. Nature 439:426–429.
Gonzalez‐Voyer A, Sxekely T, Drummond H. 2007. Why do some siblings
attack each other? Comparative analysis of aggression in avian broods.
Hanneman RA, Riddle M. 2005. Introduction to social network methods.
Riverside, CA: University of California, Riverside Available online at:
IUCN. 2011. IUCN Red list of threatened speciesVersion 2011.2. Available
online at: www.iucnredlist.org.
J Waterbird Biol 27:297–303.
James R, Croft DP, Krause J. 2009. Potential banana skins in animal social
network analysis. Behav Ecol Sociobiol 63:989–997.
King C. 1994. Management and research implications of selected behaviors
in a mixed colony of flamingos at Rotterdam Zoo. Int Zoo Yearb 33:103–
zoos. J Waterbird Biol 23:142–149.
McDonald DB. 2007. Predicting fate from early connectivity in a social
network. PNAS 104:10910–10914.
McCowan B, Anderson K, Heagarty A, Cameron A. 2008. Utility of social
Anim Behav Sci 109:396–405.
Ottenwalder JA, Woods CA, Rathburn GB, Thorbjarnarson JB. 1990.
Status of the greater flamingo in Haiti. Colon Waterbird 13:115–
Perdue B, Gaalema DE, Martin AL, Dampier SM, Maple TL. 2011. Factors
affecting aggression in a captive flock of Chilean flamingos (Phoenicopte-
rus chilensis). Zoo Biol 30:59–64.
Pritchard DJ, Fa JE, Oldfield S, Harrop SR. 2012. Bring the captive closer to
the wild: redefining the role of ex situ conservation. Oryx 46:18–23.
a tropical colony. J Waterbird Biol 26:160–165.
Réale D, Reader SM, Sol D, McDougall PT, Dingemanse NJ. 2007.
Integrating animal temperament within ecology and evolution. Biol Rev
Sandell M. 1998. Female aggression and the maintenance of monogamy:
female behaviour predicts male mating status in European Starlings. Proc
Biol Sci 265:1307–1311.
Shannon PW. 2000. Social and reproductive relationships of captive
Caribbean flamingos. J Waterbird Biol 23:173–178.
Sih A, Watters JV. 2005. The mix matters: behavioural types and group
dynamics in water striders. Behaviour 142:1417–1431.
Stevens EF, Beaumont JH, Cusson EW, Fowler J. 1992. Nesting‐behavior in
a flock of Chilean Flamingos. Zoo Biol 11:209–214.
Studer‐Thiersch A. 2000. What 19 years of observation on captive greater
flamingos suggests about adaptations to breeding under irregular
conditions. J Waterbird Biol 23:150–159.
Tuite CH. 1979. Population size, distribution, and biomass density of the
lesser flamingo in the eastern rift valley. J Appl Ecol 16:765–775.
Wey T, Blumstein DT, Shen W, Jordan F. 2008. Social network analysis of
animal behaviour: a promising tool for the study of sociality. Anim Behav
White DJ, Gersick AS, Freed‐Brown G, Snyder‐Mackler N. 2010. The
ontogeny of social skills: experimental increases in social complexity
enhance reproductive success in adult cowbirds. Anim Behav 79:385–
Wielebnowski N. 1998. Behavioral ecology and conservation biology. New
York: Oxford University Press, Inc. p 130–162.
Aggression in Captive American Flamingos