scispace - formally typeset
Search or ask a question
Journal ArticleDOI

Individual personalities predict social behaviour in wild networks of great tits (Parus major)

TL;DR: The results provide strong evidence that songbirds follow alternative social strategies related to personality, which has implications not only for the causes of social network structure but also for the strength and direction of selection on personality in natural populations.
Abstract: Social environments have an important effect on a range of ecological processes, and form a crucial component of selection. However, little is known of the link between personality, social behaviour and population structure. We combine a well-understood personality trait with large-scale social networks in wild songbirds, and show that personality underpins multiple aspects of social organisation. First, we demonstrate a relationship between network centrality and personality with ‘proactive’ (fast-exploring) individuals associating weakly with greater numbers of conspecifics and moving between flocks. Second, temporal stability of associations relates to personality: ‘reactive’ (slow-exploring) birds form synergistically stable relationships. Finally, we show that personality influences social structure, with males non-randomly distributed across groups. These results provide strong evidence that songbirds follow alternative social strategies related to personality. This has implications not only for the causes of social network structure but also for the strength and direction of selection on personality in natural populations.

Summary (3 min read)

INTRODUCTION

  • Understanding the causes and consequence of animal personalities has become one of the great challenges for recent research in evolutionary and behavioural ecology (Wolf et al. 2007; Dall et al. 2012).
  • Consistent behavioural differences between individuals have been demonstrated in multiple taxa, with some individuals repeatedly exhibiting more bold, aggressive or exploratory behaviour across a range of contexts (Sih et al. 2004).
  • An evenly spaced grid of automated feeding stations fitted with passive integrated transponder (PIT)-tag recording antennae collected ‘snap-shots’ of the composition and distribution of flocks.

Study system

  • This population has been the subject of an extensive long-term breeding survey, and there is an ongoing trapping and monitoring effort.
  • Almost all individuals in the study area are fitted with both a British Trust for Ornithology metal leg ring, and a plastic leg ring containing a uniquely identifiable PIT tag (proportion PIT-tagged estimated at over 90%, see S2 and Fig. S2).
  • While pairs of great tits defend territories over the breeding season, this breaks down into loose fission-fusion groups of unrelated individuals over autumn and winter, with roaming flocks congregating on ephemeral and patchy food sources such as beech mast (seeds of Fagus sylvatica) (Aplin et al. 2012).

Field observations

  • Adults and nestling great tits were caught in the breeding season prior to data collection (April to June 2011) and from September to November 2011, when they were aged and sexed based on plumage colour.
  • Birds were also assigned as ‘post-breeding’ adults or ‘pre-breeding’ birds (largely juveniles) based on data from previous breeding seasons (2005–2011).
  • From 2nd December 2011 until 27th February 2012, sunflower feeders were deployed at 65 locations throughout the study site, each approximately 250 m apart (Fig. S1).
  • Each feeding station had two access points each fitted with radio-frequency identification antennae and data logging hardware.
  • All feeders automatically opened from dawn to dusk on two consecutive days in every seven, scanning for PIT-tags every 16th of a second.

Behavioural assays

  • Assays of exploration behaviour in a novel environment were conducted on wild great tits that were temporarily taken into captivity at the Wytham field station over four winters (October 2009 to October 2012).
  • Most data (55%) were collected from late February to early March 2012.
  • After assays, birds were released at the site of capture.
  • Twelve types of behavioural observation were used to calculate a principal component analysis, including number of flights, flight duration, number of hops, substrates used and area explored (Quinn et al. 2009).

Statistical analysis

  • Social associations between individuals were calculated using a Gaussian mixture model that inferred group membership by detecting clusters of visits in spatio-temporal data streams (Farine et al. 2012; Psorakis et al. 2012).
  • The authors then tested if the observed pattern of associations were non-random by calculating the number of randomised networks with a higher proportion of associations and mean association strength (Whitehead 2008).
  • Social phenotype was measured using three commonly employed individual network measures; degree centrality, betweenness centrality and average association strength.
  • To avoid biasing results, all individuals that were observed in fewer than 5 of 13 sampling periods were excluded from analysis.
  • Given that each point on the surface is estimated from a large number of dyads, this test permuted the dyadic values between the two groups of data that were used to generate the same points on each of the two surfaces with respect to time lag and relative distance away from either edge of the surface (Pantazis et al. 2004).

Social associations

  • Between December 2 2011 and February 27 2012 over 3.3 million visits were recorded from 1017 individual PIT-tagged great tits observed in 26 days of data collection.
  • A social network was constructed for the whole winter period taking a ‘gambit of the group’ approach (Franks et al. 2010), inferring group membership from visitation patterns (Farine et al. 2012; Psorakis et al. 2012).
  • This remained significant when controlling for the number of spatial movements between data-loggers over the winter (LM: F1,85 = 6.3, P = 0.01), see Table S3.
  • Finally, more FE individuals were significantly more likely to move between foraging flocks, with a higher betweenness centrality (LM: F1,86 = 5.2, P = 0.02).
  • Network metrics derived at the community-level revealed the same overall relationships between personality and centrality measures (Table S4).

Temporal dynamics

  • The authors quantified the temporal stability of social relationships by estimating the lagged association rates of all post-breeding individuals with all other post-breeding individuals over the 3-month sampling period (Whitehead 2008).
  • More SE birds had a significantly higher likelihood of re-associating with other individuals, and their probability of re-association was highest with other SE birds, for which associations were maintained at a relatively high rate over time (N = 90; Fig. 2a and b).
  • In contrast, more FE birds were much less likely to re-associate, and had lower lagged association rates over the 3 month winter period (Fig. 2c and d).
  • Association rates were lowest in FE-FE interactions (Fig. 2c).
  • The effect was synergistic, with the most ephemeral relationships being between pairs of more proactive (FE) birds (Fig. 2c), and the most stable between pairs of more reactive (SE) birds (Fig. 2a; Table S5).

Social structure

  • The authors tested whether individuals of similar personality were more likely to be observed together, influencing the composition of groups and emergent social structure.
  • Preliminary analysis did, however, reveal contrasting results for mixing patterns among males and females, and the sexes were analysed separately.
  • (a) Social network where colour represents personality score ranging from most reactive (SE) phenotypes in blue to most proactive (FE) phenotypes in red; the range of the colour distribution has been slightly exaggerated at the ends of the distribution to emphasise more extreme phenotypes.
  • Grey nodes are individuals of unknown phenotype.
  • To test whether personality phenotypes were non-randomly distributed between groups, the authors calculated the kurtosis of the distribution of mean phenotype of each group.

DISCUSSION

  • Using standard behavioural assays and automated monitoring of foraging flocks, the authors show that individual-level differences in behaviour predict the frequency, stability and distribution of social associ- © 2013 John Wiley & Sons Ltd/CNRS ations in a wild songbird.
  • In particular, the authors demonstrate that individual-level variation in exploration behaviour (a proxy for the reactive-proactive axis) is associated with both social phenotype and patterns of group organisation in adult great tits.
  • Given this, there must be potentially high payoffs associated with the alternative social behaviour observed in more proactive (FE) individuals.
  • This relationship is likely to interact with ecological processes, with important implications for transmission of information and disease, and for individual variation in the acquisition of resources (Aplin et al. 2012).

Did you find this useful? Give us your feedback

Figures (2)

Content maybe subject to copyright    Report

LETTER
Individual personalities predict social behaviour in wild
networks of great tits (Parus major)
L. M. Aplin,
1,2
* D. R. Farine,
1
J.
Morand-Ferron,
1,3
E. F. Cole,
1
A.
Cockburn
2
and B. C. Sheldon
1
Abstract
Social environments have an important effect on a range of ecological processes, and form a crucial
component of selection. However, little is known of the link between personality, social behaviour and
population structure. We combine a well-understood personality trait with large-scale social networks in
wild songbirds, and show that personality underpins multiple aspects of social organisation. First, we dem-
onstrate a relationship between network centrality and personality with ‘proactive’ (fast-exploring) individu-
als associating weakly with greater numbers of conspecifics and moving between flocks. Second, temporal
stability of associations relates to personality: ‘reactive’ (slow-exploring) birds form synergistically stable
relationships. Finally, we show that personality influences social structure, with males non-randomly distrib-
uted across groups. These results provide strong evidence that songbirds follow alternative social strategies
related to personality. This has implications not only for the causes of social network structure but also for
the strength and direction of selection on personality in natural populations.
Keywords
Behavioural syndrome, Paridae, personality, social behaviour, social network theory.
Ecology Letters (2013) 16: 1365–1372
INTRODUCTION
Understanding the causes and consequence of animal personalities
has become one of the great challenges for recent research in evolu-
tionary and behavioural ecology (Wolf et al. 2007; Dall et al. 2012).
Consistent behavioural differences between individuals have been
demonstrated in multiple taxa, with some individuals repeatedly
exhibiting more bold, aggressive or exploratory behaviour across a
range of contexts (Sih et al. 2004). These consistent differences
often have a genetic basis and are likely to be subject to selection,
thereby creating the challenge of explaining how such diversity in
behavioural traits could arise and persist in natural populations
(Dingemanse et al. 2004; Dingemanse & Wolf 2010). Most current
research has concentrated on individual traits associated with varia-
tion in personality, e.g. dispersal (Quinn et al. 2011), or on dyadic
interactions, e.g. in aggression assays (Carere et al. 2005). We thus
have little understanding of the relationship between individual-level
personality traits such as exploration behaviour and social behaviour
(Webster & Ward 2011; Sih et al. 2012), or how social structure,
group dynamics and personality may interact (Krause et al. 2010).
This is a major gap, as social interactions are an important aspect
of the ecology of almost all animals, and knowledge of the social
context of personality is essential when considering potential mech-
anisms for the evolution and maintenance of personality differences
(Wolf et al. 2007; Bergmuller & Taborsky 2010; Dingemanse &
Wolf 2010).
Social network theory provides a formal framework for describing
association patterns, allowing characterisation of social structure that
integrates all levels from individual interactions to population pro-
cesses (Krause et al. 2010). If personality affects an individual’s
social behaviour, this would be expected to influence its association
patterns in the social network (Croft et al. 2009). However, the
resulting social network structure may in turn change the strength
and direction of selection on personality, if individual fitness is
dependent on the social environment (Krause et al. 2010; Wilson
et al. 2013). This patterning of social interactions may thus be
important for assessing theoretical models for the evolution of per-
sonality, most particularly selection driven by negative frequency
dependence or social niche specialisation (Wolf et al. 2007; Dinge-
manse & Wolf 2010). Under the first of these two models, payoffs
are dependent on trait frequency and network structure should thus
show a mixing of personality types (Dall et al. 2004; Johnstone &
Manica 2011), while a social niche specialisation model selection
should favour diversification or character displacement leading to
reduced social conflict (Bergmuller & Taborsky 2010).
There have been few empirical studies of the role of personality
in social networks. Most notably, Pike et al. (2008) found that cap-
tive bold sticklebacks (Gasterosteus aculeatus) had more social connec-
tions, but these interactions were more evenly spread, with shy fish
preferring to associate more strongly with fewer individuals. Conse-
quently, shoals of all bold type individuals displayed higher activity
levels. A similar result was found in captive water-striders (Aquarius
remigis), where groups of aggressive males were more active (Sih &
Watters 2005), and in shore-crabs (Carcinus maenas), where fast-
exploring individuals were more likely to make spatial movements
between groups (Tanner & Jackson 2012). Only one study has thus
far investigated the relationship between social organisation and per-
sonality in the wild, with female Trinidadian guppies (Peocilia reticulata)
more likely to be found in shoals with individuals of the same per-
sonality type (Croft et al. 2009).
1
Edward Grey Institute of Field Ornithology, University of Oxford, Oxford,
OX1 3PS, UK
2
Research School of Biology, Australian National University, Acton, 0200,
Australia
3
Department of Biology, University of Ottawa, Ottawa, K1N 6N5, Canada
*Correspondence: E-mail: lucy.aplin@anu.edu.au
© 2013 John Wiley & Sons Ltd/CNRS
Ecology Letters, (2013) 16: 1365–1372 doi: 10.1111/ele.12181

We studied personality and social behaviour in great tits, Parus
major, using the reactive-proactive personality axis common to many
vertebrate groups. This axis contrasts cautious, shy, slow-exploring
(SE) individuals with bold, aggressive, fast-exploring (FE) individu-
als; it is believed to reflect a trade-off between predator-averse
behaviour prioritising survival, and more risk-prone behaviour that
enhances productivity (Smith & Blumstein 2008; Quinn et al. 2012).
In both our population and others, an assay of exploration behav-
iour in a novel environment (performed on wild birds temporarily
taken into captivity) has been demonstrated to be a good proxy of
the reactive-proactive axis (Carere et al. 2005; Groothuis & Carere
2005; Quinn et al. 2009, 2012). Exploration behaviour has further
been shown to be repeatable (Carere et al. 2005), heritable (van
Oers et al. 2004), subject to selection (Dingemanse et al. 2004;
Quinn et al. 2009, 2011) and linked to a set of life history traits
across several populations (e.g. dispersal behaviour (Groothuis &
Carere 2005; Quinn et al. 2011)).
We use new technologies to measure social associations at a large
scale and over an extended time period in a wild wintering popula-
tion of birds. An evenly spaced grid of automated feeding stations
fitted with passive integrated transponder (PIT)-tag recording anten-
nae collected ‘snap-shots’ of the composition and distribution of
flocks. Using this spatio-temporal flocking data, we first constructed
a wild foraging social network for the entire population of 1017
individuals. Second, we used social network analysis to obtain a
measure of the social phenotype of focal individuals assayed for
personality, both at the local community and population level.
Third, we investigated the temporal stability of associations between
focal individuals over the 3 month winter flocking period. Finally,
we used two methods to ask whether the distribution of personality
types in foraging flocks was non-random, i.e. assorted, and discuss
implications for group formation and organisation. We thus present
a multi-faceted approach with complementary lines of evidence to
understand the link between individual behavioural phenotype,
social behaviour and population structure in group-living animals.
MATERIALS AND METHODS
Study system
The study was conducted on a population of great tits at Wytham
woods, Oxfordshire (51° 46’ N, 01° 20’ W). Wytham woods is a
385 ha area of broadleaf deciduous woodland, and is surrounded by
farmland (Fig. S1). This population has been the subject of an
extensive long-term breeding survey, and there is an ongoing trap-
ping and monitoring effort. Almost all individuals in the study area
are fitted with both a British Trust for Ornithology metal leg ring,
and a plastic leg ring containing a uniquely identifiable PIT tag
(proportion PIT-tagged estimated at over 90%, see S2 and Fig. S2).
While pairs of great tits defend territories over the breeding season,
this breaks down into loose fission-fusion groups of unrelated indi-
viduals over autumn and winter, with roaming flocks congregating
on ephemeral and patchy food sources such as beech mast (seeds
of Fagus sylvatica) (Aplin et al. 2012).
Field observations
Adults and nestling great tits were caught in the breeding season
prior to data collection (April to June 2011) and from September to
November 2011, when they were aged and sexed based on plumage
colour. Birds were also assigned as ‘post-breeding’ adults or
‘pre-breeding’ birds (largely juveniles) based on data from previous
breeding seasons (20052011). From 2nd December 2011 until 27th
February 2012, sunflower feeders were deployed at 65 locations
throughout the study site, each approximately 250 m apart (Fig. S1).
Each feeding station had two access points each fitted with radio-fre-
quency identification antennae and data logging hardware. All feeders
automatically opened from dawn to dusk on two consecutive days in
every seven, scanning for PIT-tags every 16th of a second. This equa-
ted to 26 days of data collection over 13 sampling periods.
Behavioural assays
Assays of exploration behaviour in a novel environment were con-
ducted on wild great tits that were temporarily taken into captivity
at the Wytham field station over four winters (October 2009 to
October 2012). Most data (55%) were collected from late February
to early March 2012. Behavioural assays have been ongoing in this
population since 2005 (Quinn et al. 2009, 2012), and we followed
existing methods, based on a design by Verbeek et al. (1994). Birds
were caught with mist-nets and housed individually overnight. On
the morning after capture, all birds were individually assayed in a
novel environment containing five artificial trees, where their move-
ments were recorded for 8 min using a handheld events recorder
(Psion Workabout, Noldus Information Technology, Nottingham,
UK) (Verbeek et al. 1994; Quinn et al. 2009). After assays, birds
were released at the site of capture. Twelve types of behavioural
observation were used to calculate a principal component analysis,
including number of flights, flight duration, number of hops, sub-
strates used and area explored (Quinn et al. 2009). PC1 described
45% of variation, and the square-root of PC1 was used in a general
linear model with individual, time of year and observation number
as fixed effects, producing a single exploration score for each indi-
vidual. In total, personality scores were collected for 221 individuals,
representing 24% of all birds observed in at least 5 of 13 field-
observation sampling periods, and 32% of all post-breeding adults.
Statistical analysis
Social associations between individuals were calculated using a
Gaussian mixture model that inferred group membership by detect-
ing clusters of visits in spatio-temporal data streams (Farine et al.
2012; Psorakis et al. 2012). This recently developed method allowed
us to detect ‘waves’ of feeding birds, without imposing arbitrary
assumptions about temporal boundaries of groups. A gambit of the
group approach (Whitehead & Dufault 1999; Franks et al. 2010)
was then used to calculate association strengths using the simple
ratio index, in which associations (or edges) are scaled between 0
(never observed in the same group) to 1 (always observed in the
same group) (Cairns & Schwager 1987). Finally, we tested whether
the observed patterns of sociality could have arisen by chance, given
spatial proximities. Permutation tests were used on the group
matrix, controlling for number of observations and group size
(Bejder et al. 1998), restricting swaps within site and within days
(Whitehead 1999, 2008). We then tested if the observed pattern of
associations were non-random by calculating the number of rando-
mised networks with a higher proportion of associations and mean
association strength (Whitehead 2008).
© 2013 John Wiley & Sons Ltd/CNRS
1366 L. M. Aplin et al. Letter

Social phenotype was measured using three commonly employed
individual network measures; degree centrality, betweenness central-
ity and average association strength. These, respectively, measure
(1) the number of other individuals with which an individual has
been observed associating with, (2) the number of shortest path
vertices to all other individuals that pass through the focal individ-
ual, important for the transmission of information and disease and
(3) the average of an individual’s edge weights, representing the
average proportion of foraging time spent with each of its associ-
ates and calculated by dividing an individual’s association strength
by its degree (Croft et al. 2008). All network analyses were con-
ducted in R Core Team (2012), using the sna and igraph packages
(Csardi & Nepusz 2006; Butts 2008).
Linear models were used to compare degree centrality, between-
ness centrality and average association strength to personality, while
adding as fixed effects the sampling periods observed, age and spa-
tial movements between data-loggers. To avoid biasing results, all
individuals that were observed in fewer than 5 of 13 sampling peri-
ods were excluded from analysis. Network communities were identi-
fied using weighted eigenvector community detection (Newman
2006). Centrality measures were then recalculated for all individuals
with network metrics derived independently from eight community-
level networks, and linear mixed models rerun with community as a
random variable. Rerunning the analysis within network communi-
ties in this way allowed the more stable local differences in social
behaviour to be isolated from the possibly confounding effects of
rare large-scale events, such as large spatial movements.
The temporal stability of relationships over time were measured
using lagged association rates, calculated as the probability of being
observed associating s days after each previous association for each
dyad [methods described in Whitehead (2008)]. We plotted the lagged
association rates as surfaces using R, and the surface calculated for
the top third of personality scores (FE) with all other individuals was
compared with the surface calculated for the bottom third of person-
ality scores (SE) with all other individuals. Areas of the surfaces sig-
nificantly different from each other were calculated using
permutation tests developed for three-dimensional surfaces. Given
that each point on the surface is estimated from a large number of
dyads, this test permuted the dyadic values between the two groups
of data that were used to generate the same (matching) points on
each of the two surfaces with respect to time lag and relative distance
away from either edge of the surface (Pantazis et al. 2004).
Finally, we gained an understanding of the relationship between
personality and social structure by calculating network assortativity,
which is a measure of the mixing patterns exhibited by individuals.
Network assortment was calculated independently for males and
females using Newman’s assortative mixing by scalar properties
(Newman 2003) in the igraph package (Csardi & Nepusz 2006), with
personality scores used as a continuous measure. Observed assort-
ment values were compared with the posterior frequency distribu-
tion calculated from 1000 node randomisations on the observed
association matrix restricted by sex. We then examined the person-
ality composition of flocks using groups inferred from the spatio-
temporal data stream. The sexes were analysed separately, and
groups including less than three individuals of known personality
score were excluded, as an accurate group mean cannot be derived
in these cases. The distribution and kurtosis score of mean group
personality phenotypes were compared to 1000 randomisations on
the group matrix.
RESULTS
Social associations
Between December 2 2011 and February 27 2012 over 3.3 million
visits were recorded from 1017 individual PIT-tagged great tits
observed in 26 days of data collection. Over 80% of individuals
were recorded in both the first and final sampling periods, indicat-
ing that winter survival was relatively high, and population turn-over
was low. Median winter range encompassed three feeding stations
with eight moves between feeders (S1). It has been suggested in
previous studies that personality may influence winter range size
(van Overveld & Matthysen 2010); however, we observed no evi-
dence for such an effect in our study (LM: F
1,203
= 0.82, P = 0.37).
There was also no relationship between personality score and num-
ber of movements between feeding stations (LM: F
1,203
= 0.1,
P = 0.83) or number of visits (LM: F
1,203
= 0.4, P = 0.53).
A social network was constructed for the whole winter period tak-
ing a ‘gambit of the group’ approach (Franks et al. 2010), inferring
group membership from visitation patterns (Farine et al. 2012; Psora-
kis et al. 2012). The temporal bounds of groups ranged from 1 s (one
visit by one individual) to 559 s; median group length 236 s. Permuta-
tion tests demonstrated that the network differed significantly from
random, even at the most local scale (P < 0.001) (Bejder et al. 1998;
Whitehead 1999, 2008). The network was also fully connected, indi-
cating a contiguous population. There was a clear difference in the
behaviour of adults that had already bred at least once previously
(‘post-breeders’; N = 285 observed in at 5 of 13 sampling periods),
and birds that were ‘pre-breeding’ (largely juveniles; N = 583
observed in at least 5 of 13 sampling periods). Pre-breeding individu-
als made much more extensive spatial movements: (GLM:
z
868
= 3.2, P = 0.001); median post-breeding total distance travelled
between feeding stations = 1.36 km, median pre-breeding distance
travelled between feeding stations = 4.18 km. There was also a differ-
ence in social behaviour, with the social associations of pre-breeding
individuals only significantly related to movement (greater movement
with higher degree centrality; LM: F
1,112
= 67.2, P < 0.00l). This was
unsurprising, as the social network was recorded over the period in
which these individuals had not fully established subsequent territories
or pair-bonds. Therefore, for the analysis of network centrality and
temporal association patterns only post-breeders were considered.
Personality and network centrality measures
Personality score in post-breeders showed a positive relationship with
degree centrality; individuals with higher exploration behaviour scores
(FE) had a larger number of social associates than individuals with
lower exploration behaviour scores (N = 90), (LM: F
1,86
= 6.1,
P = 0.0l, Fig. 1a). This remained significant when controlling for the
number of spatial movements between data-loggers over the winter
(LM: F
1,85
= 6.3, P = 0.01), see Table S3. In contrast, exploration
behaviour was negatively correlated with average association strength,
with more FE birds having on average weaker social connections than
more SE birds (LM: F
1,86
= 4.3, P = 0.04, Fig. 1b), and when con-
trolling for spatial movements (LM: F
1,85
= 4.0, P = 0.05). Finally,
more FE individuals were significantly more likely to move between
foraging flocks, with a higher betweenness centrality (LM: F
1,86
= 5.2,
P = 0.02). Three outliers exerted undue leverage on the model fit;
however, when these were removed the relationship was similar (LM:
© 2013 John Wiley & Sons Ltd/CNRS
Letter Personality and social networks in great tits, Parus major 1367

F
1,83
= 5.7, P = 0.02, Fig. 1c), and remained significant when con-
trolling for spatial movements (LM: F
1,82
= 5.0, P = 0.03) (Table S3).
To test for whether this connection between social interaction
patterns and personality occurred within ‘social cliques’ as well as at
the population level, we identified eleven cohesive network commu-
nities within the population (Newman 2006), Fig. S4. Eight of these
communities contained focal individuals. Network metrics derived
at the community-level revealed the same overall relationships
between personality and centrality measures (Table S4). Therefore,
at both the population level and within social cliques, proactive
(FE) birds were more likely to connect otherwise disparate flocks
and forage with more other individuals, but did so with a weaker
association strength.
Temporal dynamics
We quantified the temporal stability of social relationships by esti-
mating the lagged association rates of all post-breeding individuals
with all other post-breeding individuals over the 3-month sampling
period (Whitehead 2008). More SE birds had a significantly higher
likelihood of re-associating with other individuals, and their proba-
bility of re-association was highest with other SE birds, for which
associations were maintained at a relatively high rate over time
(N = 90; Fig. 2a and b). In contrast, more FE birds were much less
likely to re-associate, and had lower lagged association rates over
the 3 month winter period (Fig. 2c and d). Association rates were
lowest in FE-FE interactions (Fig. 2c). The effect was synergistic,
with the most ephemeral relationships being between pairs of more
proactive (FE) birds (Fig. 2c), and the most stable between pairs of
more reactive (SE) birds (Fig. 2a; Table S5).
Social structure
We tested whether individuals of similar personality were more likely to
be observed together, influencing the composition of groups and emer-
gent social structure. Post-breeders and pre-breeders were analysed
together, as groups were comprised of a mix of ages that did not show
strong differences in mixing patterns. Preliminary analysis did, however,
reveal contrasting results for mixing patterns among males and females,
and the sexes were analysed separately. Social structure was then inves-
(a)
(b)
(c)
Figure 1 The relationship between personality and social network position in wild great tits. (a) Social network where colour represents personality score ranging from
most reactive (SE) phenotypes in blue to most proactive (FE) phenotypes in red; the range of the colour distribution has been slightly exaggerated at the ends of the
distribution to emphasise more extreme phenotypes. Grey nodes are individuals of unknown phenotype. Size of coloured nodes represents degree. More proactive (FE)
phenotypes tend to have a larger degree centrality. (b) Average association strength decreases with personality score. (c) Positive relationship between personality and
betweenness centrality (figure is shown with 3 outliers removed; see text for analysis). Analysis was conducted on all post-breeders present in at least 5 of 13 sampling
periods (N = 90) and dashed lines represent 95% confidence intervals.
© 2013 John Wiley & Sons Ltd/CNRS
1368 L. M. Aplin et al. Letter

tigated using two complementary approaches. First, Newman’s assort-
ment measure was computed on the social network (Newman 2003).
Males tended to associate with other males of similar personality type
(N = 97; r = 0.07; P = 0.03 from 1000 node-randomisations; Fig. 3a).
Females showed no such positive assortment, with any trend in the
opposite direction to that observed in males (N = 99; r = 0.05,
P = 0.18, Fig. 3b). Second, we identified all of the discrete groups
observed at the feeding stations over winter using the spatio-temporal
data streams (N = 73 455, and generated a distribution of the mean
personality scores from these groups. To test whether personality phe-
notypes were non-randomly distributed between groups, we calculated
the kurtosis of the distribution of mean phenotype of each group. If
groups were assorted by phenotype, then the distribution of mean
group scores should be wider, resulting in a lower kurtosis score. We
then compared this score with the distribution of the kurtosis scores
from 1000 randomisations of the group matrix.
The observed distribution of personality types in groups recorded
at the data-loggers was not different from expected under random
mixing in females (no. of females = 110; Fig. 3c), with a kurtosis
score inside the distribution of kurtosis scores obtained from rando-
mised data (Fig. 3c inset). Males, however, showed a significantly
different kurtosis (no. of males = 111; Fig. 3d inset), with an
observed distribution of mean group phenotypes that fell outside of
the 95% CI of randomised data for a large part of its range (Fig. 3d).
Therefore, males within individual flocks tend to be skewed towards
particular personality types, and this supports evidence from the net-
work assortment measures that males are grouping with individuals
of similar personality. Our two alternative analyses demonstrate this
non-random mixing occurs both in the composition of short-term
flocks and over the entire winter network.
DISCUSSI ON
Using standard behavioural assays and automated monitoring of
foraging flocks, we show that individual-level differences in behav-
iour predict the frequency, stability and distribution of social associ-
(a) (c)
(b) (d)
P-value
Figure 2 Lagged association rates between individuals of differing personality. (a) Directed re-association rates between individuals with bottom third of personality scores
(SE) and all other individuals from most SE at back of plot to FE at front. Lagged association rates vary from blue (no probability of re-association between days) to red
(re-association rate of 0.2 after s days). Legend is shown at upper right. (b) Parts of plot A that significantly differ from surface in plot C; estimated from the proportion
of permuted data points where the difference between two surfaces was larger than observed. Colours show increasing significance from P = 0.05 (blue) to P < 0.001
(red); grey cells are non-significant. Legend is shown at lower right. (c) Directed re-association rates between individuals with top third of personality scores (FE) with all
other individuals, shown with SE at back to FE at front. (d) Significance surfaces for plot C.
© 2013 John Wiley & Sons Ltd/CNRS
Letter Personality and social networks in great tits, Parus major 1369

Citations
More filters
01 Jan 2012

3,692 citations

Journal ArticleDOI
TL;DR: The under‐exploited potential of experimental manipulations on social networks to address research questions is highlighted, and an overview of methods for quantifying properties of nodes and networks, as well as for testing hypotheses concerning network structure and network processes are provided.
Abstract: Summary Animal social networks are descriptions of social structure which, aside from their intrinsic interest for understanding sociality, can have significant bearing across many fields of biology. Network analysis provides a flexible toolbox for testing a broad range of hypotheses, and for describing the social system of species or populations in a quantitative and comparable manner. However, it requires careful consideration of underlying assumptions, in particular differentiating real from observed networks and controlling for inherent biases that are common in social data. We provide a practical guide for using this framework to analyse animal social systems and test hypotheses. First, we discuss key considerations when defining nodes and edges, and when designing methods for collecting data. We discuss different approaches for inferring social networks from these data and displaying them. We then provide an overview of methods for quantifying properties of nodes and networks, as well as for testing hypotheses concerning network structure and network processes. Finally, we provide information about assessing the power and accuracy of an observed network. Alongside this manuscript, we provide appendices containing background information on common programming routines and worked examples of how to perform network analysis using the r programming language. We conclude by discussing some of the major current challenges in social network analysis and interesting future directions. In particular, we highlight the under-exploited potential of experimental manipulations on social networks to address research questions.

648 citations


Cites background from "Individual personalities predict so..."

  • ...2013), how individual variation in social behaviour can drive population structure (Aplin et al. 2013; Jacoby et al. 2014; Snijders et al. 2014) and how socially transmitted quantities, such as information or disease, flow through individuals in a population (Boogert et al....

    [...]

  • ...A good example of thresholding individuals based on properties of the data is Aplin et al. (2013) who removed individuals with fewer than 100 observations as these exhibited a clear relationship between number of observations and the binary degree....

    [...]

  • ...Connor, Heithaus & Barre 2001; Wittemyer, DouglasHamilton & Getz 2005); (ii) studies of the causes and consequences of individual variation in network position – where ‘network position’ refers to the structural properties that arise as a consequence of an individual’s phenotype or patterns of sociality (e.g. McDonald 2007; Pike et al. 2008; Oh & Badyaev 2010; Aplin et al. 2013); (iii) studies of social processes and the implications of network structure for dynamics of information (e....

    [...]

  • ...The lagged association rate is useful for describing and modelling the temporal scales over which social behaviour processes operate, or for comparing how these differ between different classes of individuals (e.g. Aplin et al. 2013)....

    [...]

  • ...…properties that arise as a consequence of an individual’s phenotype or patterns of sociality (e.g. McDonald 2007; Pike et al. 2008; Oh & Badyaev 2010; Aplin et al. 2013); (iii) studies of social processes and the implications of network structure for dynamics of information (e.g. Boogert et al.…...

    [...]

Journal ArticleDOI
26 Feb 2015-Nature
TL;DR: In providing the first experimental demonstration of conformity in a wild non-primate, and of cultural norms in foraging techniques in any wild animal, the results suggest a much broader taxonomic occurrence of such an apparently complex cultural behaviour.
Abstract: In human societies, cultural norms arise when behaviours are transmitted through social networks via high-fidelity social learning. However, a paucity of experimental studies has meant that there is no comparable understanding of the process by which socially transmitted behaviours might spread and persist in animal populations. Here we show experimental evidence of the establishment of foraging traditions in a wild bird population. We introduced alternative novel foraging techniques into replicated wild sub-populations of great tits (Parus major) and used automated tracking to map the diffusion, establishment and long-term persistence of the seeded innovations. Furthermore, we used social network analysis to examine the social factors that influenced diffusion dynamics. From only two trained birds in each sub-population, the information spread rapidly through social network ties, to reach an average of 75% of individuals, with a total of 414 knowledgeable individuals performing 57,909 solutions over all replicates. The sub-populations were heavily biased towards using the technique that was originally introduced, resulting in established local traditions that were stable over two generations, despite a high population turnover. Finally, we demonstrate a strong effect of social conformity, with individuals disproportionately adopting the most frequent local variant when first acquiring an innovation, and continuing to favour social information over personal information. Cultural conformity is thought to be a key factor in the evolution of complex culture in humans. In providing the first experimental demonstration of conformity in a wild non-primate, and of cultural norms in foraging techniques in any wild animal, our results suggest a much broader taxonomic occurrence of such an apparently complex cultural behaviour.

541 citations

Journal ArticleDOI
TL;DR: It is shown that permutations of the raw observational (or ‘pre‐network’) data consistently account for underlying structure in the generated social network, and thus can reduce both type I and type II error rates.
Abstract: Null models are an important component of the social network analysis toolbox. However, their use in hypothesis testing is still not widespread. Furthermore, several different approaches for constructing null models exist, each with their relative strengths and weaknesses, and often testing different hypotheses.In this study, I highlight why null models are important for robust hypothesis testing in studies of animal social networks. Using simulated data containing a known observation bias, I test how different statistical tests and null models perform if such a bias was unknown.I show that permutations of the raw observational (or 'pre-network') data consistently account for underlying structure in the generated social network, and thus can reduce both type I and type II error rates. However, permutations of pre-network data remain relatively uncommon in animal social network analysis because they are challenging to implement for certain data types, particularly those from focal follows and GPS tracking.I explain simple routines that can easily be implemented across different types of data, and supply R code that applies each type of null model to the same simulated dataset. The R code can easily be modified to test hypotheses with empirical data. Widespread use of pre-network data permutation methods will benefit researchers by facilitating robust hypothesis testing.

312 citations


Cites background from "Individual personalities predict so..."

  • ...For example, a hypothesis might be that individuals with bold personalities have higher binary degree (more associates) in the social network (e.g. Aplin et al. 2013)....

    [...]

BookDOI
01 Jan 2016
TL;DR: In Animal Social Behaviour as discussed by the authors, the authors integrate the most up-to-date empirical and theoretical research to provide a new synthesis of the field, which is aimed at fellow researchers and postgraduate students on the topic.
Abstract: The last decade has seen a surge of interest among biologists in a range of social animal phenomena, including collective behaviour and social networks. In Animal Social Behaviour , authors Ashley Ward and Michael Webster integrate the most up-to-date empirical and theoretical research to provide a new synthesis of the field, which is aimed at fellow researchers and postgraduate students on the topic

291 citations

References
More filters
Journal Article
TL;DR: Copyright (©) 1999–2012 R Foundation for Statistical Computing; permission is granted to make and distribute verbatim copies of this manual provided the copyright notice and permission notice are preserved on all copies.
Abstract: Copyright (©) 1999–2012 R Foundation for Statistical Computing. Permission is granted to make and distribute verbatim copies of this manual provided the copyright notice and this permission notice are preserved on all copies. Permission is granted to copy and distribute modified versions of this manual under the conditions for verbatim copying, provided that the entire resulting derived work is distributed under the terms of a permission notice identical to this one. Permission is granted to copy and distribute translations of this manual into another language, under the above conditions for modified versions, except that this permission notice may be stated in a translation approved by the R Core Team.

272,030 citations


"Individual personalities predict so..." refers methods in this paper

  • ...All network analyses were conducted in R Core Team (2012), using the sna and igraph packages (Csardi & Nepusz 2006; Butts 2008)....

    [...]

Journal ArticleDOI
TL;DR: The homophily principle as mentioned in this paper states that similarity breeds connection, and that people's personal networks are homogeneous with regard to many sociodemographic, behavioral, and intrapersonal characteristics.
Abstract: Similarity breeds connection. This principle—the homophily principle—structures network ties of every type, including marriage, friendship, work, advice, support, information transfer, exchange, comembership, and other types of relationship. The result is that people's personal networks are homogeneous with regard to many sociodemographic, behavioral, and intrapersonal characteristics. Homophily limits people's social worlds in a way that has powerful implications for the information they receive, the attitudes they form, and the interactions they experience. Homophily in race and ethnicity creates the strongest divides in our personal environments, with age, religion, education, occupation, and gender following in roughly that order. Geographic propinquity, families, organizations, and isomorphic positions in social systems all create contexts in which homophilous relations form. Ties between nonsimilar individuals also dissolve at a higher rate, which sets the stage for the formation of niches (localize...

15,738 citations


"Individual personalities predict so..." refers result in this paper

  • ...However, we found no evidence for heterophily in our social network, but rather positive network assortment (i.e. homophily) among males, similar to that often observed in human personality research (McPherson et al. 2001)....

    [...]

  • ...homophily) among males, similar to that often observed in human personality research (McPherson et al. 2001)....

    [...]

01 Jan 2006
TL;DR: Platform-independent and open source igraph aims to satisfy all the requirements of a graph package while possibly remaining easy to use in interactive mode as well.
Abstract: There is no other package around that satisfies all the following requirements: •Ability to handle large graphs efficiently •Embeddable into higher level environments (like R [6] or Python [7]) •Ability to be used for quick prototyping of new algorithms (impossible with “click & play” interfaces) •Platform-independent and open source igraph aims to satisfy all these requirements while possibly remaining easy to use in interactive mode as well.

8,850 citations


"Individual personalities predict so..." refers methods in this paper

  • ...Network assortment was calculated independently for males and females using Newman’s assortative mixing by scalar properties (Newman 2003) in the igraph package (Csardi & Nepusz 2006), with personality scores used as a continuous measure....

    [...]

  • ...All network analyses were conducted in R Core Team (2012), using the sna and igraph packages (Csardi & Nepusz 2006; Butts 2008)....

    [...]

Book
30 Dec 1991
TL;DR: In this article, the development of social network analysis, tracing its origins in classical sociology and its more recent formulation in social scientific and mathematical work, is described and discussed. But it is argued that the analysis of social networks is not a purely static process.
Abstract: This paper reports on the development of social network analysis, tracing its origins in classical sociology and its more recent formulation in social scientific and mathematical work. It is argued...

6,366 citations

MonographDOI
01 Jan 2012
TL;DR: Social networks operate on many levels, from families up to the level of nations, and play a critical role in determining the way problems are solved, organizations are run, and the degree to which individuals achieve their goals.
Abstract: This book introduces the non-specialist reader to the principal ideas, nature and purpose of social network analysis. Social networks operate on many levels, from families up to the level of nations, and play a critical role in determining the way problems are solved, organizations are run, and the degree to which individuals achieve their goals. Social network theory maps these relationships between individual actors. Though relatively new on the scene it has become hugely influential across the social sciences. Assuming no prior knowledge of quantitative sociology, this book presents the key ideas in context through examples and illustrations. Using a structured approach to understanding work in this area, John Scott signposts further reading and online sources so readers can develop their knowledge and skills to become practitioners of this research method. A series of Frequently Asked Questions takes the reader through the main objections raised against social network analysis and answers the various queries that will come up once the reader has worked their way through the book.

5,439 citations

Frequently Asked Questions (2)
Q1. What have the authors contributed in "Individual personalities predict social behaviour in wild networks of great tits (parus major)" ?

The authors combine a well-understood personality trait with large-scale social networks in wild songbirds, and show that personality underpins multiple aspects of social organisation. First, the authors demonstrate a relationship between network centrality and personality with ‘ proactive ’ ( fast-exploring ) individuals associating weakly with greater numbers of conspecifics and moving between flocks. Finally, the authors show that personality influences social structure, with males non-randomly distributed across groups. This has implications not only for the causes of social network structure but also for the strength and direction of selection on personality in natural populations. 

Further research should aim to further understand the mechanisms driving emergent population structure, and attempt to establish the directionality of the relationship between social behaviour and personality traits such as exploration behaviour ( Wilson et al. 2013 ). A future challenge will be to advance the understanding of the ecology and evolution of personality by quantifying the role of personality in social networks across fluctuating spatial and temporal gradients.