A preview of this full-text is provided by Wiley.
Content available from Journal of Animal Ecology
This content is subject to copyright. Terms and conditions apply.
J Anim Ecol. 2021;90:131–142. wileyonlinelibrary.com/journal/jane
|
131© 2020 British Ecological Society
Received: 25 May 2020
|
Accepted: 7 July 2020
DOI : 10.1111/136 5-265 6.13 30 5
ANIMAL SOCIAL NETWORKS
Research Article
Testosterone-mediated behaviour shapes the emergent
properties of social networks
Roslyn Dakin1,2 | Ignacio T. Moore3 | Brent M. Horton4 | Ben J. Vernasco3 |
T. Brandt Ryder1,5
1Migratory Bird Center, Smithsonian
Conser vation Biology Institute, National
Zoological Par k, Washin gton, DC, USA
2Depar tment of Biolog y, Carleton
University, Ottawa, Ontario, Canada
3Depar tment of Biological Sciences, Virginia
Tech, Black sburg , VA, USA
4Depar tment of Biolog y, Millersville
University, Millersville, PA, USA
5Bird Conservancy of the Rockies, Fort
Collins , CO, USA
Correspondence
T. Brandt Ryd er
Email: rydert@si.edu
Roslyn Dakin
Email: roslyn.dakin@gmail.com
Funding information
NSF, Grant/Award Number: IOS 1353085;
Smithsonian Conservation Biology Institute;
Smithsonian Migrator y Bird Ce nter
Handling Editor: Damien Farin e
Abstract
1. Social networks can vary in their organization and dynamics, with implications
for ecological and evolutionary processes. Understanding the mechanisms that
drive social network dynamics requires integrating individual-level biology with
comparisons across multiple social networks.
2. Testosterone is a key mediator of vertebrate social behaviour and can influence
how individuals interact with social partners. Although the effects of testoster-
one on individual behaviour are well established, no study has examined whether
hormone-mediated behaviour can scale up to shape the emergent properties of
social networks.
3. We investigated the relationship between testosterone and social network dy-
namics in the wire-tailed manakin, a lekking bird species in which male–male so-
cial interactions form complex social networks. We used an automated proximity
system to longitudinally monitor several leks and we quantified the social network
structure at each lek. Our analysis examines three emergent properties of the
networks—social specialization (the extent to which a network is partitioned into
exclusive partnerships), network stability (the overall persistence of partnerships
through time) and behavioural assortment (the tendency for like to associate with
like). All three properties are expected to promote the evolution of cooperation.
As the predictor, we analysed the collective testosterone of males within each
social network.
4. Social networks that were composed of high-testosterone dominant males were
less specialized, less stable and had more negative behavioural assortment, after
accounting for other factors. These results support our main hypothesis that
individual-level hormone physiology can predict group-level network dynamics.
We also observed that larger leks with more interacting individuals had more
positive behavioural assortment, suggesting that small groups may constrain the
processes of homophily and behaviour-matching.
5. Overall, these results provide evidence that hormone-mediated behaviour can
shape the broader architecture of social groups. Groups with high average testos-
terone exhibit social network properties that are predicted to impede the evolu-
tion of cooperation.
132
|
Journal of Animal Ecology
DAKIN et A l.
1 | INTRODUCTION
Behavioural interactions are the foundation of social network
structures that can vary through time, among populations, and
across specie s. Net work structure plays an impor tant role in many
eco logical and evoluti onar y proces se s, inc lu ding t he spr ea d of dis-
eases (Sah, Mann, & Bansal, 2018; Stroeymey t et al., 2018), the
transmission of information and resources (Aplin, Farine, Morand-
Ferron, & Sheldon, 2012; Maldonado-Chaparro, Alarcón-Nieto,
Klarevas-Irby, & Farine, 2018) and selection on individual behaviour
(Ohtsuki, Hauert, Lieberman, & Nowak, 20 06). A major challenge
is understanding how individual-level factors, such as physiologi-
cal and behavioural mechanisms, scale up to drive the emergent
structural proper ties of social groups (Krause & Ruxton, 2002).
Linking these two levels of analysis is difficult because it requires
integrating individual-level data with repeated measures of entire
social groups (Sah, Leu, Cross, Hudson, & Bansal, 2017).
Here we use a comparison of multiple social networks through
time to investigate how hormone-mediated behaviour shapes
the higher order structure of social networks. Testosterone is a
steroid hormone that is well known for its influence on social be-
haviour and its sensitivity to changes in the social environment
(Adkins-Regan, 2005; Goymann, 2009; Wingfield, Hegner, Dufty,
& Ball, 1990). Testosterone often promotes physical aggression
and other behaviours associated with social dominance (Fuxjager
et al., 2010; Oyegbile & Marler, 20 05). Testosterone can also pro-
mote status-seeking behaviours in a non-aggressive context, in-
cluding cooperative and gregarious behaviour (Boksem et al., 2013;
Eisenegger, Haushofer, & Fehr, 2011; Ryder et al., 2020). Overall,
these hormone-signalling pathways are essential for the develop-
ment and modulation of complex behavioural phenotypes (Cohen,
Martin, Wingfield, McWilliams, & Dunne, 2012).
To investigate how testosterone is associated with social net-
work dynamics, we studied the wire-tailed manakin Pipra filicauda,
a bird species in which the males engage in coordinated displays
with each other at sites known as leks. In P. filicauda, males exhibit
two social status classes: the dominant males who hold territories
on the leks, and subordinate, ‘floater’ males who must acquire a
territor y on a lek before they can mate (Heindl, 2002; Ryder, Blake,
Parker, & Loiselle, 2011). Male social partnerships in wire-tailed
manakins can be remarkably stable and typically occur between
two unrelated males, most commonly, but not exclusively, between
a territory-owner and a floater (Dakin & Ryder, 2020; Ryder, Blake,
et al., 2011). Two features of the manakin leks make them espe-
cially well-suited to study the relationship between hormones and
group-level social structure. First, testosterone is known to af-
fect the social behaviour of individual male wire-tailed manakins
(Ryder et al., 2020). Second, these partnerships among males form
the basis of complex and dynamic social networks that are repli-
cated across leks, facilitating a comparative approach (Dakin &
Ryder, 2018, 2020). Given this background, we sought to test
whether testosterone and its effects on male behavioural pheno-
type could drive the emergent properties of the social network.
The broader function of male–male social behaviour and co-
ordinated displays in the manakin family has been the subject
of considerable study (e.g., Díaz-Muñoz, DuVal, Krakauer, &
Lacey, 2014; DuVal, 2007; McDonald, 2007; Prum, 1994; Ryder,
McDonald, Blake, Parker, & Loiselle, 2008; Ryder, Parker, Blake, &
Loiselle, 2009). One possible function is that dyadic displays may
be competitive and/or they may serve to maintain an individual's
position in a dominance hierarchy (Heindl, 2002; Prum, 1994).
Social interactions may also represent cooperative coalitions that
provide benefits to both parties (Ryder et al., 2008, 2009). The po-
tential competitive and cooperative functions of male–male social
behaviour are not mutually exclusive, and function may be context-
and/or status-specific (Ryder et al., 2008). Coordinated displays
may also be a vestige of ancestral cooperative behaviour (i.e. the
behaviour may have been directly beneficial to both parties in
the past, and it persists today, even if it no longer has adaptive
benefits; Prum, 1994). Given this background and recent evidence
that testosterone modulates male social behaviour in wire-tailed
manakins (Ryder et al., 2020), we focused this study on three
emergent proper ties of the social net works that can influence the
evolution and maintenance of cooperation (Figure 1).
The first property we examined, social specialization, captures
the exclusivity of the relationships between partners in a social
network. In the context of manakin behaviour, highly specialized
networks are well-partitioned among specific territory-owner and
floater relationships, as illustrated in Figure 1a. In humans, so-
cial specialization has been found to maximize a team's ability to
successfully per form a challenging task (Jehn & Shah, 1997). In
manakins, we expect specialization to improve the familiarit y of
social partners and the behavioural coordination of their displays.
Greater specialization is also expected to minimize conflict over
mating and territorial ascension opportunities (McDonald, 1993;
Schjelderup-Ebbe, 1922).
The second property, network stability (Figure 1b), quan-
tifies the average persistence of social partnerships through
time (Dakin & Ryder, 2020; Poisot, Canard, Mouillot, Mouquet,
& Gravel, 2012). Coordinated displays in manakins require the
synchronization of complex behaviours, and previous empir-
ical work indicates that longer partnership tenure has a posi-
tive effect on display coordination (Trainer & McDonald, 1995;
Trainer, McDonald, & Learn, 2002). Greater temporal stability of
KEYWORDS
androgens, behavioural endocrinology, collective behaviour, cooperation, dynamic networks,
social net works, testosterone
|
133
Journal of Animal Ecology
DAKIN et A l.
partnerships also increases the opportunity for familiarity and
reciprocity within a social network (Croft et al., 2006; Roberts &
Sherratt, 1998; Trivers, 1971).
The third property, behavioural assortment (Figure 1c), captures
the extent to which males interact with partners who express sim-
ilar behaviours (i.e. is like associated with like? Croft et al., 2006;
Farine, 2014). At the proximate level, positive assortment may rep-
resent the outcome of generalized reciprocity and/or partner choice
(Dakin & Ryder, 2018; Fowler & Christakis, 2010). At the ultimate
level, positive assortment has also been shown to promote the evo-
lution of cooperation (Ohtsuki et al., 2006). To quantify the overall
behavioural assortment of the manakin networks, we focused on two
correlated metrics of social behaviour within the network—‘strength’
(a male's frequency of daily social interactions) and ‘degree’ (his daily
number of social partnerships; Dakin & Ryder, 2018). We used a
composite measure of assortment that averaged the assortativity
indices of these two phenotypes.
As a potential predictor of specialization, stability, and assort-
ment, we quantified the collective testosterone of the manakin leks
( Fi gu r e 2 ; A k i n o l a, P a g e - G o ul d , M eh t a , & L u , 2 0 1 6 ). B e c au s e h o r mo n e –
behaviour relationships are status-specific in manakins and many
other species (Boksem et al., 2013; Eisenegger et al., 2011; Ryder
et al., 2020), we analyzed collective testosterone for each of the
two status classes separately. High individual testosterone levels are
associated with reduced sociality in dominant males, but increased
socialit y in subordinate males (Ryder et al., 2020). We therefore
predicted that the average (or collective) testosterone of territo-
rial males in a group would be negatively associated with its spe-
cialization, stability and assortment. In contrast, we predicted that
the collective testosterone of floater males in a group would be
FIGURE 1 Example social networks illustrating social specialization, network stability and behavioural assortment. (a) Social specialization
is measured using the bipartite form of the social net work. In the example on the left, the associations between floater males and territory-
holders are poorly partitioned, creating a network with a relatively low specialization. On the right, there is greater par titioning, such that
each floater male maintains a smaller number of associations with specific territory-holding males. (b) Network stability is measured by
evaluating the persistence of partnerships from one recording session (t1) to the next (t2). The example at the top of (b) has a lower stabilit y
than the example on the bottom. (c) Behavioural assortment measures the tendency of like to associate with like. In (c), nodes are shaded to
indicate a continuous gradient of daily behaviour (in this case, each male's average daily strength is shown). In the example network on the
left, the males tend to associate with behaviourally dissimilar partners, yielding a negative behavioural assortment. In the example on the
right, the males tend to associate with behaviourally similar partners, yielding a positive assortment. See Table 1 for additional data
Behavioural assortment
–0.52
Negative
0.66
Positive
Avg. daily behaviour
Social specialization
0.08
Low
0.83
High
(a)
(c)
FloatersTerritory
-holders
(b)
t1t2
Network stability
0.33
Low
0.88
High
134
|
Journal of Animal Ecology
DAKIN et A l.
positively associated with these three emergent properties of the
social net work.
2 | MATERIALS AND METHODS
2.1 | Study population
We studied wire-tailed manakins Pipra filicauda at the Tiputini
Biodiversity Station in Orellana Province, Ecuador (0°38′S, 76°08′W).
This population of P. filicauda ha s be e n st udied an d in d i v i d u a l s co l o u r-
banded annually since 2002 (e.g. Ryder et al., 2008, 2009). The pre-
sent study was conducted on 11 leks during peak breeding activity
(December–March) across three field seasons: 2015–2016, 2016–
2017 and 2017–2018. All research was approved by the Smithsonian
ACUC (protocols #12-23, 14-25 and 17-11) and the Ecuadorean
Ministr y of the Environment (MAE-DNB-CM-2015-0008).
2.2 | Testosterone assessment
Male manakins were captured using mist-nets on the leks as de-
scribed in (Ryder et al., 2020). We deployed up to 16 mist-nets
simu lta neo usl y at a given lek, wi th th e int ent io n of cap tur in g eve r y
male on th e lek . We rot ated the nets between lek s with th e go al of
capturing each male up to three times per field season. Each mist-
net was checked on a 30-min schedule with variation resulting
from c apture rate (i.e. multiple birds being caught on the same net
run; Ryder et al., 2020; Vernasco, Horton, Ryder, & Moore, 2019).
The amount of time a male spends in the net has a subtle, but sig-
nificant, negative effect on his circulating testosterone (Vernasco
et al., 2019). Therefore, we used video monitoring to determine
the duration of time that each bird was in the mist-net prior to
blood sampling, so that it could be accounted for in further analy-
ses (M = 17.5 min, SD = 10.3 min, range = 1–72 min). Following re-
moval of a male from the mist-net, a small blood sample (<125 μl)
FIGURE 2 Collective testosterone of manakin social networks. (a) Circulating testosterone varies among individual males. This graph
shows repeated measures from 210 individual male manakins, sorted along the x-axis by mean testosterone, a standardized measure of a
male's average circulating testosterone level (Ryder et al., 2020). Opacit y is used to denote the two male status classes (with subordinate
floaters coloured semi-transparent, and dominant territory-holders coloured opaque). Colour ramping is used to denote each individual's
hormone phenotype. (b–d) These data were used to define collective testosterone in the present study. Node colours indicate an individual's
mean testosterone following the scale in (a). The collective testosterone of each social network is calculated as the average of the individual
hormone phenotypes, weighted by strength, and is shown below each example. Collective testosterone was determined for each status
class separately. See Table 1 for additional data
–2
–1
0
1
2
Log T (ng/ml)
–2 –1 01–2 –1 01 –2 –1 01
–2 –1 01
Mean T
Territory-holders
Floaters
(a)
(b) (c) (d)
Collective testosterone
High T maleLow T male
High T group
Low T group
|
135
Journal of Animal Ecology
DAKIN et A l.
was taken from the brachial vein and stored on ice prior to being
centrifuged at ~11,000 g for 5 min, as described in previous stud-
ies (Ryder et al., 2020; Ryder, Horton, & Moore, 2011; Vernasco
et al., 2019). Plasma volume was measured to the nearest 0.25 μl
and stored in 0.75 ml of 100% ethanol (Goymann, Schwabl,
Trappschuh, & Hau, 2007). Testosterone was double extracted
from the plasma using dichloromethane. Following extraction, a
direct radioimmunoassay was used to measure the total plasma
androgen concentration (ng/ml) adjusted by the extraction ef-
ficiency and plasma volume of each sample (Eikenaar, Whitham,
Komdeur, van der Velde, & Moore, 2011; Ryder, Horton, et al.,
2011). Hormone assays were conducted annually, and the detec-
tion limits were 0.12, 0.08 and 0.09 ng/ml for 2015–2016, 2016–
2017 and 2017–2018, respectively; any sample that fell below
the assay-specific limit of detection was assigned that limit as its
testosterone concentration as a most conser vative estimate. The
extraction efficiency for all samples was between 62% and 73%,
and the intr a-assay coefficients of variation were 6.6%, 11.6% and
9.2% for 2015–2016, 2016–2017 and 2017–2018, respectively;
the inter-assay coefficient of variation was 19.5%.
2.3 | Behavioural recording
We used an automated data-logging system to monitor male–
male interactions on the display territories of the leks (Dakin &
Ryder, 2018; Ryder et al., 2020; Ryder, Hor ton, van den Tillaart,
Morales, & Moore, 2012). The territories on these leks are specifi-
cally used for male–male coordinated social displays (as described
in Schwar tz & Snow, 1978). At the beginning of each field season,
male manakins were outfitted with coded nano-tags (NTQB-2,
Lotek Wireless; 0.35 g). The tags transmitted a unique VHF signal
ping once per 20 s for 3 months. In total, 296 tag deployment s were
performed on 180 individuals (mean 1.7 field seasons per male ± SD
0.7), 178 of whom also had hormone data (mean number of hormone
samples per male = 3 ± SD 1. 5). Approximately 10 days (±SD 7) af ter
tagging and sampling was completed at a given lek, a proximity data-
logger (SRX-DL800, Lotek Wireless) was deployed within each dis-
play territory at the lek to record tagged males within a detection
radius of 30 m (a distance that corresponds to the typical diameter
of a manakin display territory; Dakin & Ryder, 2018; Heindl, 2002).
Proximity recording sessions ran from 06:00 to 16:00 for ~6 con-
secutive days (±SD 1 day) and were performed ~3 times per field
season at a given lek. Occasionally, the length of a recording session
was extended due to extenuating circumstances such as inclement
weather. The recording sessions were scheduled to be distributed
evenly throughout each field season at each lek, to minimize any
confounding seasonal effects. Each recording session represents an
observation of the social network at a given lek. In total, we con-
ducted 86 recording sessions (29,760 data-logger hours) represent-
ing repeated measures of the social activity at 11 leks during three
field seasons (see Figure S1 in the Suppor ting Information for addi-
tional details on the sampling regime).
Prior to data-logger deployment, each territory was also ob-
served on non-recording days to identify the territory-holder based
on his colour bands, following previous studies (Ryder et al., 2008,
2009). These status assignments were subsequently verified in the
proximity data.
2.4 | Data processing and statistical analysis
All computational and statistical analyses were performed in r (R Core
Team, 2018). Network illustrations were made using the i gr aph
package (Csardi & Nepusz, 2006).
2.5 | Social interactions
Social networks were constructed by first defining interactions be-
tween two males that occurred on the display territories. To do this,
a computational algorithm was used to identif y joint detections,
wherein two males were located at the same display territory within
a pre-defined spatial and temporal threshold (Dakin & Ryder, 2018;
Ryder et al., 20 08, 2012). For the temporal threshold, the two males
had to occur <45 s apart. This temporal threshold was chosen to
allow for the fact that each tag pinged with a 20-s pulse rate, such
that overlapping individuals could have up to a 40-s gap between
their respective pings. For the spatial threshold, the two males had
to have a difference in received signal strength values (ΔRSSI) < 10 .
This threshold corresponded to a typical distance between 0 and
5 m apar t in a ground-truthing experiment (Dakin & Ryder, 2018).
Hence, according to our definition, a social interaction is initiated
only after males come within this approximate spatial threshold.
We chose this spatial threshold because it is close enough to permit
visual and acoustic contact during t ypical social behaviours (such as
those described in Schwart z & Snow, 1978). After completing our
study, we also performed a sensitivity analysis to verify that our
main results were robust to alternative spatial threshold definitions
(see Supporting Information for details).
Any repeated co-occurrence of the two males within 5 min was
considered to be part of the same social interaction, but after a
gap of ≥5 min, it was considered to be a new interaction between
those two males. The average duration of social interactions de-
fined by this method was 5.1 min ± SD 13.3, 8.8 min ± SD 20.1 and
7.0 min ± SD 20.5 in the three respective field seasons, further in-
dicating that these were sustained social interactions, rather than
random encounters.
In total, we identified 36,885 social interactions over the three
field seasons of study. These interactions were used to define a
weighted social network for each lek recording session. The nodes in
the network were the individual males, and the edges were weighted
by the number of social interactions between each pair of males.
An earlier validation study compared social interactions that were
detected by the proximity system with those that were directly ob-
served for 11 males (Ryder et al., 2012), and confirmed that all of
136
|
Journal of Animal Ecology
DAKIN et A l.
the interactions detected by the proximity system were also directly
observed.
2.6 | Null model validation of the social networks
Two broad classes of methods have been described for build-
ing social net works: (a) networks that are built based purely on
the proximity of animals unrelated to their behavioural context
(‘gambit of the group’), and (b) networks that are built based on
specific behavioural criteria that are directly obser ved (Crof t,
Madden, Franks, & James, 2011; Farine, 2015; Franks, Ruxton,
& James, 2010). Although the interactions in our study were
not directly observed, they were recorded at specialized display
perches that have a known function in male–male social interac-
tions (Ryder, Blake, et al., 2011; Schwartz & Snow, 1978). Hence,
they do not qualify as the gambit of the group. To further dem-
onstrate this point, we performed a pre-network permutation of
the raw data to determine whether the observed network edges
occurred more often than expected by chance (Farine, 2017).
This analysis was highly conservative in that it prese rved key fea-
tures of the data including male visit rates to specific territories
within each recording session and lek; it is described in detail in
the Supporting Information and Figure S2. The results demon-
strated that 95% of the observed network edges had a greater
edge weight than expected under stringent permutation condi-
tions. Moreover, these preferred edges had edge weights that
were 10-fold to 50-fold greater than expected by chance under
the se strin gent condition s (F igure S2). This prov ides an ad ditiona l
validation to our approach , bec ause it indica te s that the o bs er ved
network edges were nearly always preferred, even relative to
other possible interactions within the same lek. These results are
also consistent with a previous validation study indicating that
these methods capture male–male coalition partnerships that are
directly observed (Ryder et al., 2012).
Although pre-network data permutations are sometimes used to
derive adjusted association indices, or to prune networks prior to fur-
ther analysis, we did not take this approach for several reasons. First,
the statistical rarity of a relationship in our system does not a priori
define the importance of any one social interaction, especially given
that all of the interactions occurred on display territories with spe-
cialized function. A single interaction between two rarely interacting
males may have been highly consequential (e.g. if one of those males
had a highly influential hormone–behavioural phenotype). Conversely,
a single interaction between two frequently interac ting males may
have been relatively unimportant. It would be unwarranted to assume
that partnership rarit y indicates the impor tance of any single interac-
tion in this context. Second, and perhaps more importantly, the goal of
this study was a comparison across multiple networks. Some networks
are genuinely less preferential, and more random, than others. If we
modified the network edges based on permutation-based indices, it
would disproportionately prune the truly random networks, introduc-
ing a source of bias that would be contrar y to our main goal. Hence, all
further analyses are based on networks where the edge weight s are
given by the observed number of interactions, as this method is most
appropriate for our study system and aims. Below, we also describe a
separate node-label permutation that provided an additional check on
our final statistical analyses.
2.7 | Social specialization
To quantify social specialization, we sought a metric that would
capture the extent to which a network was partitioned into ex-
clusive social relationships (as opposed to a network made up of
non-specific or non-exclusive par tnerships). To do this, we used
a network metric of specialization that is commonly used in com-
munity ecology called
H′
2
(Blüthgen, Menzel, & Blüthgen, 2006).
An advantage of
H′
2
is that it is standardized against a theoretical
maximum, based on the overall activity levels of different nodes
and Shannon entropy (Blüthgen et al., 2006); this makes it pos-
sible to compare the extent of specialization across different bi-
partite net works in a standardized way. To apply this metric to
our manakin data, we converted each lek's social net work into
its bipartite adjacency matrix (Figure 1a), with floaters along one
axis, and territory-holders on the other, and then calculated social
specialization as
H′
2
using the b ipar tit e package (Dormann, Fruend,
& Gruber, 2019). Higher values of specialization indicate that the
network is well-partitioned (i.e. made up of exclusive relation-
ships), as illustrated in Figure 1a. We chose to focus on floater-
territorial specialization because these two social classes are
well-defined and floater-territorial partnerships tend to be the
most common in this species (Figure S3; see also Ryder, Blake,
et al., 2011).
Our measure social specialization at the network level,
H′
2
,
can also be related to the exclusivity of social partnerships at
the individual level, as used in (Dakin & Ryder, 2018; Edenbrow
et al., 2011; Sih, Hanser, & McHugh, 2009). All else being equal, a
highly specialized network is expected to be made up of individu-
als who are relatively more exclusive and/or more important to-
ward their partners, sensu (Dakin & Ryder, 2018; Sih et al., 2009).
Because
H′
2
has th e prop e r tie s desc rib ed in the pre vio us par agr aph ,
it is more appropriate as the network-level metric of social spe cial-
ization. Note that it is possible, and perhaps even common, for the
edge weights in a highly specialized network to be relatively invari-
ant if the network is not fully connected, as shown in the example
manakin networks in Figure 1a. Hence, specialization based on
H′
2
is not in principle related to the network-average coefficient of
variation of each individual's edge weights (Maldonado-Chaparro
et al., 2018).
2.8 | Network stability
We define network stability as the average persistence of net-
work edges through time (Dakin & Ryder, 2020). To quantify the
|
137
Journal of Animal Ecology
DAKIN et A l.
stability of manakin networks, we compared each lek's social net-
work from a given recording session to its subsequent recording
session within the same field season (Figure 1b). Network stabil-
ity was then calculated as the number of male–male partnerships
(binary network edges) shared by both time point s divided by the
number of partnerships at either time point (Dakin & Ryder, 2020).
Higher values of stabil it y in dicate greater persis tence of socia l re-
lationships within the network, independent of any changes in the
representation of particular males (nodes) (Poisot et al., 2012). To
focus this measure on the persistence of strong relationships, we
computed network stability of par tnerships that occurred at least
six times following (Dakin & Ryder, 2020). The threshold of six was
chosen because it corresponds to an average rate of one social
inter ac tion per day in our data. We conducted a n addi ti on al sen si-
tiv it y a nalysis to verify th at alter native thresholds for the st abil it y
calculation (greater or less than six) did not change our main re-
sults (see Supporting Information for details). Previous work using
this metric of s tability has shown that the wire-tailed manakin so-
cial networks are far more stable than expected by chance (Dakin
& Ryder, 2020).
2.9 | Behavioural assortment
Assortment refers to the extent to which individuals associate
with similar partners (Figure 1c); it can be due to partner choice
(homophily), shared environments or the social transmission of
behaviour (Dakin & Ryder, 2018). Assortment was quantified
using Newman's assor tativity index, which is a correlation coef-
ficient for the statistical association among linked nodes within
a network. It ranges from –1 (a negative association), through 0
(no association), and up to +1 (a positive association). To quan-
tify the assortment of social behaviours, we first determined
the daily frequency of two behaviours for each male—his num-
ber of social interactions per day (strength), and his number of
unique social partnerships per day (degree). Strength and degree
are both repeatable measures of a male's social behaviour in
our study population (Dakin & Ryder, 2018). We used the aver-
age log-transformed values of each male's strength and degree
within the recording session, and then calculated the assortativ-
ity coefficient for the entire social network using the algorithm
for weighted networks in the ass ort net package (Farine, 2016).
Because assortativit y values for strength and degree were highly
correlated (Pearson's r = 0.78, p < 0.0001, n = 86 networks), we
took the average of these two values as the measure of over-
all behavioural assortment within the social network. Note that
we used log-transformed values of strength and degree because
these two variables are strongly positively skewed (Dakin &
Ryder, 2018; Ryder et al., 2020), and assor tativity is based on a
Pearson's correlation. Finally, we also computed the assortativity
of the two discrete status classes (floater and territory-holder), to
ensure that our analysis of behavioural assortment was not solely
driven by status-assortment.
2.10 | Collective testosterone
To understand how hormones might predict network properties, we
derived a measure of collective testosterone of each social network.
This was based on the hormonal trait that was the best predictor of
social behaviour in our previous study, referred to as ‘mean testos-
terone’ (Ryder et al., 2020). A male's mean testosterone is his aver-
age residual circulating testosterone. It is calculated using a linear
regression of log-transformed testosterone to statistically account
for the effects of field season, Julian date, time of day when cap-
tured and duration of restraint, all of which may influence point es-
timates of a male's baseline hormone level (Vernasco et al., 2019).
Hence, mean testosterone represents a standardized measure of a
male's circulating testosterone, independent of his capture condi-
tions (Figure 2a). Next, to determine collective testosterone, we took
the average mean testosterone for each social network, weighted by
the interaction frequency (strength) of the males within the network
(Figure 2b). Collec tive testosterone is thus a group-level character-
istic that is weighted towards the males that made the greatest con-
tribution to group social structure. In other words, networks with
low collective testosterone are made up of mostly low-testosterone
individuals, whereas networks with high collective testosterone are
made up of mostly high-testosterone individuals. We calculated
collective testosterone separately for each status class, because
the effects of hormones on social behaviour are status-dependent
(Boksem et al., 2013; Eisenegger et al., 2011; Ryder et al., 2020).
2.11 | Statistical analysis
To evaluate the hypothesis that collective testosterone predicts
network proper ties, we analysed mixed-effects models of the social
network properties in the lme4 package (Bates, Maechler, Bolker, &
Walker, 2018). The three response variables were social specializa-
tion, network stability and behavioural assortment (Figure 1). We
used Akaike's Information Criterion (AIC) to compare four candidate
models for each response variable, as follows: (a) collective testos-
terone of territory-holders + collective testosterone of floaters;
(b) collective testosterone of territory-holders; (c) collective tes-
tosterone of floaters and (d) no testosterone predictors. All of the
models included additional fixed effects to account for field season
(a categorical variable with three levels), the average Julian date of
the recording session, the average number of recorded hours per
territor y and the size of the social network (number of individuals),
as well as a random effect of lek to account for repeated measures.
Model comparison was performed on models fit with maxi-
mum likelihood, and the best-fit models were re-fit using re-
stricted estimation of maximum likelihood (REML) to derive
parameter estimates (Zuur, Ieno, Walker, Saveliev, & Smith, 2009).
We used the lme rtest package to compute p-values for parameter
estimates in the mixed-effects models based on Satterthwaite's
method (Kuznetsova, Brockhoff, & Christensen, 2018). We also
report Nakagawa and Schielzeth's
R2
LMM
values as an estimate of
138
|
Journal of Animal Ecology
DAKIN et A l.
effect size (Nakagawa & Schielzeth, 2013). We verified that all
models met the assumptions of linear regression analyses. First,
we checke d that th e Pearson resid ua ls met th e assumption of nor-
mality. We visually inspected the partial residual plots for each
fixed and random effect, to confirm that there were no outliers
or depar tures from the assumption of homoscedasticit y. To check
for multicollinearity, we used the pe rfo rma nce package (Lüdecke,
Makowski, & Waggoner, 2019) to calculate variance inflation fac-
tors (VIFs), and we verified that all VIFs were bet ween 1 and 1.8.
2.12 | Sample sizes and exclusions
In two field seasons (2016–2017 and 2017–2018), we performed an
experiment as part of a separate study to test the influence of ex-
perimentally elevated testosterone on individuals (n = 5 individuals in
2016–2017 and n = 4 in 2017–2018; Ryder et al., 2020). The results
of that experiment demonstrated that elevated testosterone caused
a temporary decrease in the frequenc y and the number of social part-
nerships in the altered males (Ryder et al., 2020). It is important to
note that this experiment was not designed to test emergent proper-
ties at the network level, because it was conducted on a limited scale
with only one or two individuals temporarily altered within each lek.
We therefore excluded the six post-manipulation networks from the
main analysis in this study. We verified that when we included these
manipulated leks, all of our main conclusions were unchanged.
After excluding the six post-manipulation observations, 80 of
the original 86 recording sessions remained. Table 1 summarizes the
sample sizes for the network-level analyses. In 10 cases, specializa-
tion could not be calculated because the bipartite network did not
have sufficient data to determine
H′
2
. Stability could not be calcu-
lated in 26 cases, when either the recording session occurred at the
end of a field season, or when there were insufficient partnerships
that met the criteria for the stability calculation.
2.13 | Node-label permutation analysis
To evaluate the possibilit y that our result s could be influenced
by other properties of the leks that were independent of hor-
monal traits, we also performed a statistical permutation of the
post-network data (Farine, 2017). The purpose of this analysis
was to verify that the results were driven by (and sensitive to) the
relative contribution and social position of different males. To do
this, we performed a statistical permutation that randomized the
node labels (male IDs) within each of the social networks, retaining
network topology, and leaving each male's testosterone traits un-
changed. Hence, this analysis preserved which males were present
in which recording session, but it randomized the relative position
and contribution of each male. After generating 1,000 of these
node-label permutation datasets, we recalculated specialization,
stability, and assortment, and then refit the top models from our
mixed-model analysis. We then compared the slope estimates from
the observed data to those derived from 1,000 node-label permu-
tations. As a one-sided p-value, we calculated the proportion of
slope estimates from the node-label permutations that were more
negative than the corresponding estimate from the observed data.
3 | RESULTS
The lek social networks had 14 males on average, but there was
considerable variation among leks ranging from 3 to 43 individuals
(Figure S1). Additional descriptive statistics are provided in Table 1.
Although network size has a lognormal distribution (Figure S1), all
other emergent properties of the social structure were approxi-
mately normally distributed (Figure S4).
We found that the collective testosterone of territorial males
could predict all three emergent properties of the social networks
(specialization, stabilit y and assortment). The leks with greater
representation of high-testosterone territorial males were less
specialized, less stable and more negatively assorted (Figure 3; all
p ≤ 0.0 3 in mixed-ef fe ct s models). The slop e co ef fi cient s for these
three relationships were also all greater than expected under a
node-label permutation (inset panels, Figure 3; all p < 0.02). Model
selection results for the observed data indicated considerable
uncertainty in the best-fit model for each of the three network
properties analysed (Table S1). However, the collective testoster-
one of territory-holders was a significant predictor in all of the
best-supported models (Table S2). In contrast, the collec tive tes-
tosterone of floater males was not a significant predictor of net-
work properties in any of the best-fit models (Table S2).
TABLE 1 Descriptive statistics for manakin social networks. Means and SD are provided for each field season. The bottom row provides
the sample sizes for the number of social networks analysed (nobs) at each lek (Nlek). Additional data are provided in Figures S1 and S4
Field season
Network
size (# birds)
Social
specialization
Network
stability
Behavioural
assortment
Collec tive T
(te rr. )
Collec tive T
(floa.)
15 –1 6 14.5 (8.9) 0.50 (0.22) 0.43 (0.25) 0.01 (0.36) 0.44 (0.42) −0.31 (0.56)
16 –1 7 17.8 (10. 5) 0.30 (0.24) 0.47 (0.20) 0.06 (0.30) 0.49 (0.24) −0.22 (0.37)
17–1 8 10.3 (4.5) 0.42 (0.30) 0.42 (0.27) 0.02 (0.37) 0.43 (0.40) −0.70 (0.46)
Sample size nobs = 80
Nlek = 11
nobs = 70
Nlek = 11
nobs = 54
Nlek = 11
nobs = 80
Nlek = 11
nobs = 80
Nlek = 11
nobs = 80
Nlek = 11
|
139
Journal of Animal Ecology
DAKIN et A l.
We did not detect any significant effects of Julian date on net-
work proper ties within our study period, but we did observe some
significant year-to-year differences (e.g. social specialization and
behavioural assortment in Table S2). Assor tment was the only prop-
erty that was significantly related to network size and recording
effort (Table S2). All else being equal, behavioural assortment was
more positive in larger networks, and it was more negative in net-
works that had longer recording sessions.
R2
LMM(m)
provides an estimate of the proportion of variance ex-
plained by the fixed effects in a model (Table S1). The
R2
LMM(m)
for
assortment in the best-fit model was 0.52. This indicates that about
52% the variation in behavioural assortment could be explained by
the combined associations with collective testosterone, network
size, sampling effort and annual variation (Table S2). The
R2
LMM(m)
for
stability was 0.12, indicating that collective testosterone and the
other predictors (network size, sampling effor t, year) could explain
about 12% the variation in that metric (Table S2). Finally, for special-
ization, the
R2
LMM(m)
indicated that the combined effects of testoster-
one and these other sources of variation together explained about
20% of the variation (Table S2).
4 | DISCUSSION
The collective testosterone of the dominant, territor y-holding males
within a lek was associated with multiple emergent properties of
the social network (Figure 3). Variation in collective testosterone
is a function of both the number of high-testosterone males and
their frequency of social interactions. Our results indicate that the
hormone-mediated behaviour of these individuals may affect all
three social network properties of specialization, stability and as-
sortment. This indicates that the effects of testosterone on domi-
nant males may mediate an extended phenotype with the power to
shape social structure (Dawkins, 1982). In contrast, the collective
testosterone of floater males was not significantly associated with
any of the emergent network properties (Table S2).
How can the relationship between hormone levels and network
properties be explained in terms of individual mechanisms? Given that
testosterone has antagonistic effects on the sociality of territorial
males (Ryder et al., 2020; Vernasco, Horton, Moore, & Ryder, 2020),
we hypothesize that the behaviour of high-testosterone individuals
can cause several features of the social organization to break down.
Our previous results showed that high-testosterone dominant indi-
viduals have a reduced abilit y to attract and maintain social partners
(Ryder et al., 2020). We propose that this weakening of social rela-
tionships may cause floater males to prospect elsewhere for new
partners, negatively affecting both the stability and specialization
of the lek network as a whole. Likewise, high-testosterone domi-
nant males may inhibit the processes of social contagion, reciprocit y
and/or behavioural matching that can cause positive behavioural
assortment (Dakin & Ryder, 2018). Recent studies have found that
in other social animals, sparse and specialized social networks can
be associated with fitness benefits (Stroeymeyt et al., 2018); hence,
a breakdown to this organization may incur costs (Maldonado-
Chaparro et al., 2018). Testing our proposed mechanism for the link
between hormone-mediated behaviour and network dynamics will
FIGURE 3 Collective testosterone predicts the structure
of the social network. The net works with higher testosterone
dominant males were less specialized, less stable over time, and
had more negative behavioural assortment. Each plot shows the
partial residuals from a statistical analysis that also accounts for
field season, the average Julian date of the recording session, the
number of recording hours, network size and lek identity. Because
floater and territorial males can differ in behaviour, the analysis of
behavioural assortment also accounted for status assortment within
each network. Different symbols are used to indicate repeated
measures of 11 different leks, coloured according to the collective
testosterone scale in Figure 2. Inset panels show the result s of
node-label permutation tests. In each case, the slope in the best-
fit model (solid vertical line) is significantly more negative than
expected based on the distribution of permuted estimates (grey
distribution, dotted line is at 0). See Tables S1 and S2 for
additional data
0.00.5 1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.00.5 1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.00.5 1.0
–0.4
–0.2
0.0
0.2
0.4
0.6
Tower
P75
P2.5-800
P2-200
P1050
MQ3600
HL
H3-1000
GL
Chichico
Al Rio
Network
Collective testosterone
(territory-holders)
Social specialization
(residual)
Behavioural assortmen
t
(residual)
Network stability
(residual)
140
|
Journal of Animal Ecology
DAKIN et A l.
require direct observation of the individual social behaviours that
occur within dyads, and how these behaviours change through time.
Because our current data cannot assess the fine-scale valence of
social interactions, further studies are needed that combine direct
observation with high-throughput data on social network dynamics.
We included several additional parameters in our analyses to ac-
count for social network size and sampling effort. Although it was
not one of our main hypotheses, we noted that network size was
positively associated with behavioural assortment (Table S2). In other
words, males were more likely to associate with behaviourally similar
partners in larger leks, whereas they were more likely to associate
with dissimilar par tners in smaller leks. Effects of group size on as-
sortment have been noted in a few other studies, although the form
of this relationship varies (Griffiths & Magurran, 1997; Ilmarinen,
Vainikainen, Verkasalo, & Lönnqvist, 2017; McDonald, Farine, Foster,
& Biernaskie, 2017). In manakins and other lekking systems, larger
leks are known to have heightened display activity and higher fe-
male visitation rates (Durães, Loiselle, Parker, & Blake, 2009; Lank
& Smith, 1992). This raises the possibility that social facilitation and
he igh t ene d ac tiv ity may be assoc iat e d wit h incr e ase d ho m oph i ly an d /
or behavioural matching (either through partner choice, contagion,
or reciprocity). Another plausible explanation is that smaller social
groups may constrain these behavioural processes, by making it more
difficult to find or match a suitable partner. This hypothesis could be
explored in experiments on captive systems and simulation models.
Behavioural assortment was also more positive in networks that
were recorded for less time in our study (in other words, leks where
males associated with behaviourally similar par tners tended to be
recorded for fewer days). Recording time was designed to be approx-
imately even among leks (Figure S1), with the most common reason
for an exte nde d re cor di ng du ratio n being incl em ent wea the r that ex-
tended the recording session. Therefore, we speculate that inclem-
ent weather may have reduced the amount of behavioural matching
(assortment) while also affecting recording time. Although we did
not collect data on weather at each lek as this was not our main goal,
the question of how inclement weather affects group-level social
dynamics is an interesting one that merits further study.
Conducting experimental tests of causation at the level of whole
social networks remains a major challenge in ecology (James, Croft, &
Krause, 2009; Pinter-Wollman et al., 2014). Although our node-label
permutation analysis provides evidence that our results are not
merely due to structural differences among networks, we cannot
rule out the possibility that high-testosterone individuals chose to
participate in certain networks due to other factors that may also
influence the emergent properties of the network (e.g. environmen-
tal quality and/or female activity). It is important to note that in this
system, as in many other wild animals, controlled experimental ma-
nipulations of the broader social network structure are not yet pos-
sible (Akinola et al., 2016; Zyphur, Narayanan, Koh, & Koh, 2009).
Nevertheless, our data here indicate that the increased prevalence
of dominant, high-testosterone individuals can predict changes in
social dynamics and subsequent higher order network structure.
These findings establish that hormone–behaviour relationships are
not limited to one individual, but instead hormones have population-
level consequences (McClintock, 1981; Robinson, 1992).
Increasing evidence demonstrates that the structure and stability
of so cia l net wor k s is of ten associat ed with benef it s —and cos ts —during
foraging, breeding and disease outbreaks (Maldonado-Chaparro
et al., 2018; Riehl & Strong, 2018; Silk et al., 2010; Stroeymeyt
et al., 2018). Given the widespread influence of steroid hormones
on social interactions across ver tebrates (Adkins-Regan, 2005), we
expect that collective testosterone will be broadly associated with
social net work properties in other systems. The direction of these ef-
fects may depend on the context of the behavioural interactions that
form the social network, and whether these interactions are primarily
cooperative or competitive in nature. Taken together with previous
studies, we propose that testosterone-mediated behaviour can alter
social network dynamics in ways that often impede the evolution of
stable social relationships and cooperation.
ACKNOWLEDGEMENTS
We thank Camilo Alfonso, Brian Evans, David and Consuelo Romo,
Kelly Swing, Diego Mosquera, Gabriela Vinueza and the Tiputini
Biodiversity Station of the Universidad San Francisco de Quito.
Funding was provided by the National Science Foundation (NSF)
IOS 1353085 and the Smithsonian Migrator y Bird Center.
AUTHORS' CONTRIBUTIONS
R.D., I.T.M., B.M.H. and T.B.R. designed the study; I.T.M., B.M.H.,
B.J.V. and T.B.R. collected the data; R.D., I.T.M. and T.B.R. analysed
the data; R.D. and T.B.R . wrote the manuscript. All authors edited
the manuscript.
DATA AVAILAB ILITY STATE MEN T
All data and R scripts necessar y to reproduce this study are available
for download at: https://doi.org/10.608 4/m9.figsh are.83210 36.v1
(Dakin, Moore, Hor ton, Vernasco, & Ryder, 2020).
ORCID
Roslyn Dakin https://orcid.org/0000-0002-3140-3975
T. Brandt Ryder https://orcid.org/0000-0002-5517-6607
REFERENCES
Adkins-Regan, E. (2005). Hormones and animal social behavior. Princeton,
NJ: Princeton University Press.
Akinola, M., Page-Gould, E., Mehta, P. H., & Lu, J. G. (2016). Collective
hormonal profiles predict group performance. Proceedings of the
National Academy of Sciences of the United States of America, 113,
9774–9779. https://doi.org/10.1073/pnas.16034 43113
Aplin, L. M., Farine, D. R., Morand-Ferron, J., & Sheldon, B. C . (2012).
Social networks predict patch discovery in a wild population of
songbirds. Proceedings of the Royal Society B: Biological Sciences, 279,
4199–4205 . ht tps://doi.o rg/10.1098/rspb.2 012.1591
Bates, D., Maechler, M., Bolker, B., & Walker, S. (2018). lme4 1.1-18-1:
Linear mixed-effects models using ‘Eigen’ and S4. Vienna, Austria:
Comprehensive R Archive Network.
Blüthgen, N., Menzel, F., & Blüthgen, N. (2006). Measuring specialization
in species interaction networks. BMC Ecology, 6, 9.
|
141
Journal of Animal Ecology
DAKIN et A l.
Bo k sem , M. A . S ., Me hta , P. H. , Va n den Be r gh , B., van So n, V., Traut man n,
S. T., Roelofs, K ., … Sanfey, A . G . (2013). Testos terone inhibits trust
but promotes reciprocity. Psychological Science, 24, 2306–2314.
https://doi.org/10.1177/09567 97613 495063
Cohen, A. A., Martin, L . B., Wing field, J. C., McWilliams, S. R., & Dunne,
J. A. (2012). Physiologic al regulatory networks: Ecological roles and
evolutionary constraints. Trends in Ecology & Evolution, 27, 428–435.
https://doi.org/10.1016/j.tree.2012.04.008
Croft, D. P., James, R., Thomas, P. O. R., Hathaway, C., Mawdsley, D.,
Laland, K . N., & Krause, J. (2006). Social structure and co-operative
interac tions in a wild population of guppies (Poecilia reticulata).
Behavioral Ecology and Sociobiology, 59, 644–650. https://doi.
org/10.1007/s0026 5-005-0091-y
Croft, D. P., Madden, J. R ., Franks, D. W., & James, R. (2011). Hypothesis
testing in animal social networks. Trends in Ecology & Evolution, 26,
502–507. https://doi.org/10.1016/j.tree.2011.05.012
Csardi, G., & Nepusz, T. (2006). The igraph software package for complex
network research. InterJournal, Complex Systems, 1695(5), 1–9.
Dakin , R., Moore, I. T., Horton, B. M., Vernasco, B. J., & Ryder, T. B. (2020).
Data from: Supplementary Materials for: Testosterone-mediated be-
havior shapes the emergent properties of social networks. Dataset.
figshare, https://doi.org/10.6084/m9.figsh are.83210 36.v1
Dakin, R., & Ryder, T. B. (2018). Dynamic network partnerships and so-
cial contagion drive cooperation. Proceedings of the Royal Society B:
Biological Sciences, 285, 20181973.
Dakin, R., & Ryder, T. B. (2020). Reciprocity and behavioral heterogeneity
govern the stability of social networks. Proceedings of the National
Academy of Sciences of the United States of America, 117, 2993–2999.
https ://doi.or g/10.1073/pnas.19132 8 4117
Dawkins, R. (1982). The extended phenot ype: The gene as the unit of selec-
tion. San Francisco, CA: W. H. Freeman and Company.
Díaz-Muñoz, S. L., DuVal, E. H., Krakauer, A . H., & Lacey, E. A. (2014).
Cooper ating to compete: Altruism, sexual selection and causes of
male reproductive cooperation. Animal Behaviour, 88, 67–78. https://
doi.org/10.1016/j.anbeh av.2013.11.008
Dormann, C. F., Fruend, J., & Gruber, B. (2019). bipar tite 2.13: Visualising
bipartite networks and calculating some (ecological) indices. Vienna,
Austria: Comprehensive R Archive Network.
Durães, R., Loiselle, B. A., Parker, P. G., & Blake, J. G. (2009). Female mate
choice across spatial scales: Influence of lek and male attributes on
mating success of blue-crowned manakins. Proceedings of the Royal
Society B: Biological Sciences, 276, 1875–1881.
DuVal, E. H. (20 07). Adaptive advantages of cooperative cour tship for
subordinate male lance-tailed manakins. The Ameri can Naturalist, 169,
423–432. https://doi.org/10.1086/512137
Edenbrow, M., Darden, S. K., Ramnarine, I. W., Evans, J. P., James, R., &
Croft, D. P. (2011). Environmental effec ts on social interaction networks
and male reproductive behaviour in guppies, Poecilia reticulata. Animal
Behaviour, 81, 551–558. https://doi.org/10.1016/j.anbeh av.2010.11.026
Eikenaar, C., Whitham, M., Komdeur, J., van der Velde, M., & Moore, I. T.
(2011). Endogenous testosterone is not associated with the trade-
off bet ween paternal and mating effor t. Behavioral Ecology, 22, 6 01–
608. https://doi.org/10.1093/behec o/arr030
Eiseneg ger, C., Haushofer, J., & Fehr, E. (2011). The role of testoster-
one in social interaction. Trends in Cognitive Sciences, 15, 263–271.
https://doi.org/10.1016/j.tics.2011.04.008
Farine, D. R. (2014). Measuring phenotypic assor tment in animal social
networks: Weighted associations are more robust than binary edges.
Animal Behaviour, 89, 141–153. https://doi.org/10.1016/j.anbeh av.
2014.01.001
Farine, D. R. (2015). Proximity as a proxy for interactions: Issues of scale
in social net work analysis. Animal Behaviour, 104, e1–e5. https://doi.
org/10.1016/j.anbeh av.2014.11.019
Farine, D. R. (2016). assor tnet 0.12: calculate the assor tativity coeff icient of
weighted and binary networks.
Farine, D. R. (2017). A guide to null models for animal social network
analysis. Methods in Ecology and Evolution, 8, 1309–1320. https://doi.
org /10.1111/2041-210X .12772
Fowler, J. H., & Christakis, N. A . (2010). Cooperative behavior cascades
in human social networks. Proceedings of the National Academy of
Sciences of the United States of America, 107, 5334–5338. https://doi.
org /10.1073/pnas.09131 49107
Franks, D. W., Rux ton, G. D., & James, R. (2010). Sampling animal as-
sociation net works with the gambit of the group. Behavioral Ecology
and Sociobiology, 64, 493–5 03 . ht tp s: //doi.org /10.10 07/s0 026 5 -
009- 0865-8
Fuxjager, M. J., Forbes-Lorman, R. M., Coss, D. J., Auger, C. J., Auger,
A. P., & Marler, C. A. (2010). Winning territorial disputes selectively
enhances androgen sensitivity in neural pathways related to moti-
vation and social aggression. Proceedings of the National Academy of
Sciences of the United States of America, 107, 12393–12398. https://
doi.org/10.1073/pnas.10013 94107
Goymann, W. (2009). Social modulation of androgens in male birds.
General and Comparative Endocrinology, 16 3, 149–157. https://doi.
org/10.1016/j.ygcen.2008.11.027
Goym ann , W., Schwabl, I., Trappschuh, M., & Hau , M. (2 007 ). Use of eth-
anol for preserving steroid and indoleamine hormones in bird plasma.
General and Comparative Endocrinology, 150, 191–195. https://doi.
org/10.1016/j.ygcen.2006.09.014
Griffiths, S. W., & Magurran, A. E. (1997). Schooling preferences for famil-
iar fish vary with group size in a wild guppy population. Proceedings of
the Royal Society of London. Series B: Biological Sciences, 264, 547–551.
Heindl, M. (2002). Social organization on leks of the wire-tailed manakin
in Southern Venezuela. The Condor, 104, 772–779. https://doi.
org /10.1093/cond o r/104.4.772
Ilmarinen, V.-J., Vainikainen, M.-P., Verkasalo, M. J., & Lönnqvist, J.-E.
(2017). Homophilous friendship assortment based on personality
trait s and cog nitive ab il it y in middle chi ld ho od : The moderating ef fect
of peer net work size. European Journal of Personality, 31, 20 8–219.
James, R., Croft, D. P., & Krause, J. (2009). Potential banana skins in an-
imal social network analysis. Behavioral Ecology and Sociobiology, 63,
989–997. https://doi.org/10.1007/s0026 5-009-0742-5
Jehn, K. A., & Shah, P. P. (1997). Interpersonal relationships and task per-
formance: An examination of mediation processes in friendship and
acquaintance groups. Journal of Personality and Social Psycholog y, 72,
775–790. htt ps://doi.org/10.10 37/002 2-3514.72.4.7 75
Krause, J., & Ruxton, G. D. (2002). Living in groups. Oxford, UK: Oxford
University Press.
Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2018). lmerT-
est 3.0-1: Tests in linear mixed ef fects models. Vienna, Austria:
Comprehensive R Archive Network.
Lank, D. B., & Smith, C . M. (1992). Females prefer larger leks: Field ex-
periments with ruffs (Philomachus pugnax). Behavioral Ecology and
Sociobiology, 30, 323–329. https://doi.org/10.1007/BF001 70598
Lüdecke, D., Makowski, D., & Wag goner, P. (2019). performance 0.4.2:
Assessment of regression models performance. Vienna, Austria: Com-
pre hensive R Archive Network .
Maldonado-Chaparro, A. A., Alarcón-Nieto, G., Klarevas-Irby, J. A., &
Farine, D. R. (2018). Experimental disturbances reveal group-level
costs of social instability. Proceedings of the Royal Society B: Biological
Sciences, 285, 20181577.
McClintock, M. K. (1981). Social control of the ovarian cycle and the function
of estrous synchrony. Integrative and Comparative Biology, 21, 243–256.
McDonald, D. B. (1993). Delayed plumage maturation and orderly queues
for status: A manakin mannequin experiment. Ethology, 94, 31–45.
https://doi.org/10.1111/j.1439-0310.1993.tb005 45.x
McDonald, D. B. (2007). Predicting fate from early connec tivity in a social
network. Proceeding s of the National Acad emy of Sciences of the Un ited
States of America, 104, 10910–10914. https://doi.org/10.1073/
pnas.07011 59104
142
|
Journal of Animal Ecology
DAKIN et A l.
McDonald, G. C., Farine, D. R., Foster, K. R., & Biernaskie, J. M. (2017).
Assor tment and the analysis of natural selection on social traits.
Evolution, 71, 2693–2702. https://doi.org/10.1111/evo.13365
Nakagawa, S., & Schielzeth, H. (2013). A general and simple method for
obtaining R2 from generalized linear mixed-effects models. Methods
in Ecology and Evolution, 4, 133–142.
Ohtsuki, H., Hauert , C ., Lieberman, E., & Nowak, M. A. (20 06). A sim-
ple rule for the evolution of cooperation on graphs and social
networks. Nature, 441 , 502–505. https://doi.org/10.1038/natur
e04605
Oyegbile, T. O., & Marler, C. A. (20 05). Winning f ights elevates testos-
terone levels in California mice and enhances future ability to win
fights. Hormones and Behavior, 48, 259–267. https ://doi.org/10.1016/
j.yhbeh.2005.04.007
Pinter-Wollman, N., Hobson, E. A., Smith, J. E., Edelman, A. J., Shizuka,
D., de Silva, S., … McDonald, D. B. (2014). The dynamics of animal
social networks: Analytical, conceptual, and theoretical advances.
Behavioral Ecology, 25, 242–255. https://doi.org/10.1093/behec o/
art0 47
Poisot, T., Canard, E., Mouillot, D., Mouquet, N., & Gravel, D. (2012).
The dissimilarity of species interaction networks. Ecology Letters, 15,
1353–1361. https: //doi.org/10.1111/el e.120 02
Prum, R. O. (1994). Phylogenetic analysis of the evolution of alternative
social behavior in the manakins (Aves: Pipridae). Evolution, 48, 1657–
1675. h tt ps://doi.or g/10.1111/ j.1558-5646.1994.tb 02 2 03 .x
R Core Team. (2018). R 3.5.1: A language and environment for statistical
computing. Vienna, Austria: R Foundation for Statistical Computing.
Riehl, C., & Strong, M. J. (2018). Stable social relationships between
unrelated females increase individual fitness in a cooperative
bird. Proceedings of the Royal Society B: Biological Sciences, 285,
20180130.
Roberts, G., & Sherratt, T. N. (1998). Development of cooperative rela-
tionships through increasing investment. Nature, 394, 175. https://
doi.org/10.1038/28160
Robinson, G. E. (1992). Regulation of division of labor in insect societies.
Annual Review of Entomology, 37, 637–665. https://doi.org/10.1146/
annur ev.en.37.010192.003225
Ryder, T. B., Blake, J. G., Parker, P. G., & Loiselle, B. A . (2011). The com-
position, stabilit y, and kinship of reproductive coalitions in a lekking
bird. Behavioral Ecology, 22, 282–290. https://doi.org/10.1093/behec
o/arq213
Ryder, T. B., Dakin, R., Vernasco, B. J., Evans, B. S., Horton, B. M., &
Moore, I. T. (2020). Testosterone modulates status-specific patterns
of cooperation in a social network. The American Naturalis t, 195(1),
82–94. https://doi.org/10.1086/70 6236
Ryder, T. B., Horton, B. M., & Moore, I. T. (2011). Unders tanding tes-
tosterone variation in a tropic al lek-breeding bird. Biology Letters, 7,
506–509. https://doi.org/10.1098/rsbl.2010.1219
Ryder, T. B., Hor ton, B. M., van den Tillaar t, M., Morales, J. D. D., &
Moore, I. T. (2012). Proximity data-loggers increase the quantity and
quality of social network data. Biology Letters, 8, 917–920. https://
doi.org/10.1098/rsbl.2012.0536
Ryder, T. B., McDonald, D. B., Blake, J. G., Parker, P. G., & Loiselle, B. A .
(2008). Social networks in the lek-mating wire-tailed manakin (Pipra
filicauda). Proceedings of the Royal Society B: Biological Sciences, 275,
1367–1374.
Ryder, T. B., Parker, P. G., Blake, J. G., & Loiselle, B. A. (2009). It takes
two to tango: Reproductive skew and social correlates of male mat-
ing success in a lek-breeding bird. Proceedings of the Royal Society B:
Biological Sciences, 276 , 2377–2384.
Sah, P., Leu, S. T., Cross, P. C., Hudson, P. J., & Bansal, S. (2017). Unraveling
the disease consequences and mechanisms of modular structure
in animal social networks. Proceedings of the National Academy of
Sciences of the United States of America, 114 , 4165–4170. https://doi.
org/10.1073/pnas.1613 6 16114
Sah, P., Mann, J., & Bansal, S. (2018). Disease implic ations of animal social
network structure: A synthesis across social systems. Journal of Animal
Ecology, 87, 546–558. https://doi.org/10.1111/1365-2656.12786
Schjelderup-Ebbe, T. (1922). Beiträge zur sozialpsychologie des haushuhns
[Observation on the social psychology of domestic fowls]. Zeitschrift
für Psychologie und Physiologie der Sinnesorgane. Abt. 1. Zeitschrift für
Psychologie, 88, 225–252.
Schwartz, P., & Snow, D. W. (1978). Display and related behavior of the
wire-tailed manakin. Living Bird, 17, 51–78.
Sih, A., Hanser, S. F., & McHugh, K. A. (2009). Social network theory: New
insights and issues for behavioral ecologists. Behavioral Ecology and
Sociobiology, 63, 975–988. https://doi.org/10.1007/s0026 5- 009-0725-6
Silk, J. B., Beehner, J. C., Bergman, T. J., Crockford, C., Engh, A. L.,
Moscovice, L. R., … Cheney, D. L. (2010). Strong and consistent social
bonds enhance the longevity of female baboons. Current Biology, 20,
1359–1361. https://doi.org/10.1016/j.cub.2010.05.067
Stroeymeyt, N., Grasse, A. V., Crespi, A., Mersch, D. P., Cremer, S., &
Keller, L. (2018). Social network plasticity decreases disease trans-
mission in a eusocial insect. Science, 362, 941–945. https://doi.
org/10.1126/scien ce.aat4793
Trainer, J. M., & McDonald, D. B. (1995). Singing performance, frequency
matching and courtship success of long-tailed manakins (Chiroxiphia
linearis). Behavioral Ecology and Sociobiology, 37, 249–254 . ht tp s://doi.
org /10.1007/BF001 7740 4
Trainer, J. M., McDonald, D. B., & Learn, W. A. (2002). The development of
coordinated singing in cooperatively displaying long-tailed manakins.
Behavioral Ecology, 13 , 65–69. https://doi.org/10.1093/behec o/13.1.65
Trivers, R. L. (1971). The evolution of reciprocal altruism. The Quarterly
Review of Biology, 46, 35–57. https://doi.org/10.1086/406755
Vernasco, B. J., Horton, B. M., Moore, I. T., & Ryder, T. B. (2020). Reduced
cooperative behavior as a cost of high testosterone in a lekking pas-
serine bird. Behavioral Ecology, 31, 401–410. https://doi.org/10.10 93/
behec o/arz201
Vernasco, B. J., Horton, B. M., Ryder, T. B., & Moore, I. T. (2019). Sampling
baseline androgens in free-living passerines: Methodological con-
siderations and solutions. General and Comparative Endocrinology,
Endocrinology of Neotropical Vertebrates, 273, 202–208. https://doi.
org/10.1016/j.ygcen.2018.07.017
Wingfield, J. C., Hegner, R. E., Dufty, A. M., & Ball, G. F. (1990). The
“challenge hypothesis”: Theoretical implications for patterns of tes-
tosterone secretion, mating systems, and breeding strategies. The
American Naturalist, 136, 829–84 6. https://doi.org/10.1086/285134
Zuur, A ., Ieno, E. N., Walker, N., Saveliev, A . A., & Smith, G. M. (2009).
Mixed effects models and extensions in ecolog y with R. New York, NY:
Springer Science & Business Media.
Zyphur, M. J., Narayanan, J., Koh, G., & Koh, D. (2009). Testosterone–
status mismatch lowers collective effic acy in groups: Evidence from a
slope-as-predictor multilevel structural equation model. Organizational
Behavior and Human Decision Processes, Biological Basis of Business, 110,
70–79. https://doi.org/10.1016/j.obhdp.20 09.05.004
SUPPORTING INFORMATION
Additional supporting information may be found online in the
Supporting Information section.
How to cite this article: Dakin R, Moore IT, Horton BM,
Vernasco BJ, Ryder TB. Testosterone-mediated behaviour
shapes the emergent properties of social networks. J Anim
Ecol. 2021;90:131–142. ht tps://doi. or g/10.1111/1365 -
2656.13305
