Heterogeneous structure in mixed-species corvid ﬂocks in ﬂight
Jolle W. Jolles
, Andrew J. King
, Andrea Manica
, Alex Thornton
Department of Psychology, University of Cambridge, Cambridge, U.K.
Department of Zoology, University of Cambridge, Cambridge, U.K.
College of Science, Swansea University, Swansea, U.K.
Centre for Ecology and Conservation, University of Exeter, Cornwall Campus, Penryn, U.K.
Received 5 November 2012
Initial acceptance 2 January 2013
Final acceptance 16 January 2013
Available online 1 March 2013
MS. number: 12-00829
Flocks of birds in ﬂight represent a striking example of collective behaviour. Models of self-organization
suggest that repeated interactions among individuals following simple rules can generate the complex
patterns and coordinated movements exhibited by ﬂocks. However, such models often assume that
individuals are identical and interchangeable, and fail to account for individual differences and social
relationships among group members. Here, we show that heterogeneity resulting from species differ-
ences and social structure can affect ﬂock spatial dynamics. Using high-resolution photographs of mixed
ﬂocks of jackdaws, Corvus monedula, and rooks, Corvus frugilegus, we show that birds preferentially
associated with conspeciﬁcs and that, like high-ranking members of single-species groups, the larger and
more socially dominant rooks positioned themselves near the leading edge of ﬂocks. Neighbouring birds
showed closer directional alignment if they were of the same species, and neighbouring jackdaws in
particular ﬂew very close to one another. Moreover, birds of both species often ﬂew especially close to
a single same-species neighbour, probably reﬂecting the monogamous pair bonds that characterize these
corvid social systems. Together, our ﬁndings demonstrate that the characteristics of individuals and their
social systems are likely to result in preferential associations that critically inﬂuence ﬂock structure.
Ó2013 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
How do large aggregations of individuals, each of which may
differ in its preferred outcome, coordinate their movements? The
spectacular displays of ﬂocking birds led the naturalist Edmund
Selous (1931) to postulate a role for ‘thought transference’, but
recent advances have begun to unravel the mysteries of collective
movement without appealing to the supernatural (Couzin & Krause
2003;Conradt & Roper 2005;Sumpter 2006). Models of self-
organizing systems suggest that repeated interactions among in-
dividuals following simple rules can generate complex patterns and
coordinated group movements. Models of agents following simple
rules of (1) long-range attraction to group members, (2) short-
range repulsion and (3) alignment between close neighbours
have generated realistic representations of collective animal
movements (reviewed in Sumpter 2006;Petit & Bon 2010). How-
ever, empirical veriﬁcation of their assumptions remains scarce and
largely conﬁned to model systems such as starlings, Sturnus vulgaris
(e.g. Ballerini et al. 2008a,b;Hemelrijk & Hildenbrandt 2011).
Mathematical models of self-organization commonly assume
that individuals are identical, independently interacting agents
(Vicsek & Zafeiris 2012), but this is unlikely to be realistic (Sumpter
2006;Petit & Bon 2010). Group members often mix associatively
according to a variety of morphological and physiological factors
such as sex, size and energetic state (reviewed in Krause & Ruxton
2002) and species’social systems have been shown to inﬂuence
the spatial distribution of individuals in a variety of contexts (Krause
1993;King et al. 2008;Jacobs et al. 2011). However, studies of col-
lective behaviour seldom consider the impact of such heterogeneity
upon the spatial dynamics of ﬂocks, or the rules of interaction un-
derlying their coordination. Recent studies suggest that these im-
pacts may be critical. Harcourt et al. (2009), for example,
demonstrated that individual differences have substantial impacts
on coordination rules in pairs of sticklebacks, Gasterosteus aculeatus,
while Nagy et al. (2010) identiﬁed a hierarchical structure in homing
pigeon ﬂocks, Columba livia domestica, with key individuals con-
tributing disproportionately to the group’s movement decisions.
Mixed-species ﬂocks provide excellent opportunities for
empirical investigations into the impacts of heterogeneity on ﬂock
structure. Species differences may generate nonrandom organiza-
tions of individuals within ﬂocks (Latta & Wunderle 1996), while
members of larger or more dominant species may play a pivotal
role in leading group movements (Goodale & Beauchamp 2010).
Mixed-species ﬂocks are an important form of social organization
for birds worldwide, and an extensive literature suggests that
*Correspondence: A. Thornton, Centre for Ecology and Conservation, University
of Exeter, Cornwall Campus, Penryn, Cornwall TR10 9EZ, U.K.
E-mail address: firstname.lastname@example.org (A. Thornton).
Contents lists available at SciVerse ScienceDirect
journal homepage: www.elsevier.com/locate/anbehav
0003-3472/$38.00 Ó2013 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Animal Behaviour 85 (2013) 743e750
species differences are reﬂected in the spatial structure and
movements of foraging groups. For instance, certain species may
play a disproportionate role in ﬂock formation and cohesion, while
species that are particularly vulnerable to predation often follow
and exploit the vigilance of heterospeciﬁcs (Sridhar et al. 2009;
Goodale & Beauchamp 2010). However, as research has focused on
foraging interactions, very little is known about the structure of
mixed-species ﬂocks in ﬂight. Analyses of such aerial ﬂocks can
provide important insights into the interaction rules governing
Using high-resolution photographs of jackdaws, Corvus mon-
edula, and rooks, Corvus frugilegus, in ﬂight, we examined the ef-
fects of species differences and social systems on mixed-species
ﬂocks. Jackdaws and rooks spend a large portion of the year for-
aging and roosting together in large groups. During the winter,
ﬂocks of up to 1000 or so individuals leave their foraging grounds
and ﬂy to preroost trees before aggregating in a single ﬂock num-
bering in the thousands above the roost where they spend the night
(Coombs 1961). The social system of both species centres around
long-term monogamous pair bonds (Emery et al. 2007), but rooks
are larger and dominant in foraging interactions and access to
roosting sites (Lockie 1956;Coombs 1961). Thus, these ﬂocks are
neither homogeneous nor composed of anonymous individuals,
and so provide an ideal system to investigate how heterogeneity
(speciﬁcally species differences and social relationships) can
mediate the movement rules that individuals adopt, and hence
inﬂuence ﬂock structure.
We assumed that ﬂocking rooks and jackdaws would not
interact in an identical manner to all neighbours (cf. Nagy et al.
2010), and that this would be reﬂected in ﬂock structure. Speciﬁ-
cally, we predicted (1) that individuals would associate preferen-
tially with conspeciﬁcs and (2) that, like high-ranking members of
single-species groups (King et al. 2009;Nagy et al. 2010), the so-
cially dominant rooks would position themselves near the leading
edge of ﬂocks. If birds preferentially interact with speciﬁc in-
dividuals, then we predicted (3) greater proximity and alignment
among conspeciﬁc than heterospeciﬁcneighbours. Alone, such
assortment and alignment could simply reﬂect differing aero-
dynamic or morphological constraints between the two species,
rather than differential reactions depending on neighbours’spe-
cies. However, such constraints would not be expected to result in
the occurrence of discrete dyads of individuals within ﬂocks. Con-
sequently, our ﬁnal prediction (4) was that birds should show
increased proximity to a single same-species social partner, which
is likely to reﬂect the monogamous pair-bonded societies of these
corvids (Emery et al. 2007).
We photographed corvid ﬂocks moving to and from pre-
roosting sites before combining in a single large ﬂock above the
roost (sunset 45 min), between 19 October 2011 and 8 February
2012 in an area of approximately 0.3 km
village of Madingley, Cambridgeshire, U.K (see Appendix Fig. A1).
Photographs were taken perpendicular to the ﬂocks’ﬂight di-
rection at a distance of approximately 100e300 m, from different
locations throughout each evening so as to avoid pseudor-
eplication from repeated shots of the same ﬂock. The number of
different ﬂocks photographed per evening ranged from one to 11
(mean ¼3.1 0.8). We used a Canon EOS 7D digital SLR camera
with a Canon EF 100e40 0 mm f/4.5e5.6 L IS lens. We set the
camera to Auto Focus with Av exposure mode, with photos taken
in RAW and settings adjusted to maximize distinguishability
between the features of jackdaws and rooks. The drive mode was
set to high-speed continuous shooting (8 frames/s), allowing us
to capture sets of consecutive images from the front, middle and
back thirds of ﬂocks (hereafter ‘ﬂock section’).
Photo Editing and Species Identiﬁcation
Jackdaws and rooks are visually distinctive. Jackdaws are
smaller, with a short, black bill, grey nape, blue/grey eyes and
a wide tail in ﬂight, while rooks are larger with entirely black
plumage, a long, bald beak, dark eyes, a relatively narrow tail and
primary wing feathers typically splayed in a ﬁnger-like fashion in
ﬂight. To maximize clarity and enable species identiﬁcation of as
many birds as possible, we edited all photographs using the Adobe
Photoshop Camera Raw plugin (Adobe Systems, San Jose, CA,
U.S.A.). We then identiﬁed rooks and jackdaws from the edited
photographs based on body size, head shape, beak shape, wing
shape and tail shape. From a total of 1211 photographs, editing
allowed us to identify the species identity of >95% of birds in 144
photographs. For analysis, we excluded photographs in which the
total ﬂock size was less than 20 (as small ﬂocks would not permit
analyses based on seven nearest neighbours in front, middle and
back; see below) and the few images from ﬂocks consisting entirely
of a single species. This ﬁnal data set contained a total of 115
photographs from 44 ﬂocks (N¼44 from the front and middle and
N¼27 from the back of ﬂocks; each ﬂock was assigned a unique
Flock Identity). Following editing, we merged all photos of front,
middle and back sections to form one larger image of the whole
ﬂock (‘ﬂock image’). We counted the total number of birds in each
ﬂock image as a proxy for total ﬂock size and noted the proportion
of rooks in each ﬂock. As birds were not individually identiﬁable in
ﬂight, it is possible that the same ﬂock may have been photo-
graphed on different evenings. However, ﬂock sizes varied sub-
stantially, from 21 to 638 individuals, and there were only three
instances (from a total of 44 ﬂocks) where we photographed ﬂocks
of the same size over different evenings. Our collection of photo-
graphs is therefore likely to represent a large sample of different
Alignment and Proximity of Neighbours
To examine the alignment and proximity of neighbours, we
randomly selected four focal birds from each ﬂock section (front,
middle and back), noting their species and that of their nearest
neighbours. We chose four focal birds because (1) this allowed us
to have several representatives from each ﬂock section but (2)
the number of focal birds per section was sufﬁciently low that we
could ensure focal birds would never be nearest neighbours to
each other, which would result in pseudoreplication. If two
randomly selected birds were both nearest to one other, they
were only considered in the analysis once and a new bird was
randomly selected. We determined the distance between the
midpoints of neighbouring birds in jackdaw lengths (based on
the average body length of seven randomly selected jackdaws in
the ﬂock). To determine the directional alignment between
neighbours, we used the ‘ruler tool’in Photoshop CS5, by drag-
ging the tool from the midpoint of the tail and beyond the
midpoint of the head of each bird, thus providing the angle of the
line through the body, relative to horizontal in the photograph.
The difference between the angles of neighbouring birds was
used as a measure of alignment. Our estimates of distances and
alignment between neighbours necessarily involve some error as
they rely on two-dimensional representations of the true three-
dimensional structure of ﬂocks. However, while these errors
introduce some noise into the data, they generate no directional
J. W. Jolles et al. / Animal Behaviour 85 (2013) 743e750744
biases. Our estimates are therefore likely to provide robust yet
conservative measures of the true degree of structure in ﬂocks.
Data were analysed in Genstat 14.1 (VSN International, Hemel
Hempstead, U.K.) using linear mixed models (LMM) or generalized
linear mixed models (GLMM) for normal and non-normal data,
respectively, with ﬂock identity nested in date as a random term to
control for repeated measures in all cases. Initially, all probable
explanatory variables were entered into the model. All possible
interactions between them were investigated and terms were
sequentially dropped until the minimal model contained only
terms whose elimination would signiﬁcantly reduce the explan-
atory power of the model. Wald statistics and probability values for
signiﬁcant terms were derived from the minimal model containing
only signiﬁcant terms, while values for nonsigniﬁcant terms were
obtained by adding each term individually to the minimal model
(Crawley 2002). The residuals for all models were visually inspec-
ted to ensure homogeneity of variance, normality of error and
linearity. All results with P<0.05 are reported as signiﬁcant. Means
are quoted SE throughout. Post hoc analyses of differences be-
tween levels within categorical variables (e.g. front, middle, back)
were conducted by sequentially excluding each level from (G)LMM
analyses to enable comparisons of the remaining category levels.
Tables of results for all multifactorial analyses including all effect
sizes and SEs are in the Appendix.
Preferential associations by species
To test whether the birds showed preferential associations by
species (prediction 1) we randomly selected four focal birds per
ﬂock section and ran a GLMM with binary response term (1,0)
testing the probability that a focal bird’s neighbour was a jackdaw.
Explanatory terms were focal bird species and the proportion of
rooks in the ﬂock.
Positional differences by species
To compare positional differences between the species in ﬂocks,
we randomly selected one focal bird in each ﬂock section, noting its
species and that of its seven nearest neighbours. We used seven
neighbours because previous research indicates that individuals in
starling ﬂocks interact with a ﬁxed number of six or seven neigh-
bours (Ballerini et al. 2008a). Unlike the analyses of associations,
distances and alignments between neighbours, there was no need to
restrict analyses to four birds per ﬂock section to avoid pseudor-
eplication. To test whether rooks ﬂew disproportionately near the
leading edge of ﬂocks (prediction 2) we used a GLMM with a bino-
mial response term (number of rooks out of the total of eight birds)
and ﬂock section (front, middle or back) as an explanatory variable.
Flock size, the proportion of rooks, month (to control for possible
seasonal variation) and time relative to sunset (because individuals’
motivation to reach preferred sites within the roost may increase as
night approaches) were ﬁtted as additional variables.
Proximity and alignment between neighbours
To test whether distance and alignment differed between con-
speciﬁcand heterospeciﬁc neighbours (prediction 3) we noted the
distance (in jackdaw lengths) and directional alignment between
focal birds (four per ﬂock section) and their nearest neighbours (see
Appendix). We then ran two LMMs with neighbour distance and
neighbour alignment as response terms and dyad type (jackdaws,
rooks or mixed) as our variable of interest, along with ﬂock section
(front, middle, back) and ﬂock size. Distances were square-root
transformed and alignments were normalized for analysis using
aBoxeCox power transformation.
Identiﬁcation of discrete dyads within ﬂocks
Field observations and visual inspection of photographs indi-
cated that jackdaws and rooks commonly ﬂy in discrete dyads
within ﬂocks (Coombs 1961 reported similar observations). To
conﬁrm this, we used a custom-made script written in R (www.R-
project.org) to measure the distance between all individuals (in
jackdaw or rook lengths, from the midpoint of each bird) and their
seven nearest same-species neighbours in a selection of nine ﬂock
section photographs. In very dense ﬂocks, even discrete dyads
would tend to ﬂy near other dyads. As an illustrative sample, we
therefore chose photographs of ﬂock sections in which the density
was sufﬁciently low to allow us to identify dyads clearly. The pho-
tographs used to examine jackdaw and rook dyads were not always
the same, as some images contained insufﬁcient rooks. Using the
neighbour distance measurements, we conducted the following
(1) Categorization of discrete dyads and triads. We deﬁned
discrete dyads as same-species neighbours whose interindividual
distance was less than half the distance to the second closest
neighbour. This conservative measure is likely to underestimate the
true frequency of discrete dyads in ﬂocks as discrete dyads could
nevertheless ﬂy close to other discrete dyads. We also investigated
the occurrence of same-species triads of birds, deﬁned as cases in
which the nearest-neighbour distances between three birds were
all less than half the distance to the fourth neighbour. Triads may
occur among corvids when unpaired individuals (either adult birds
that had lost their partner or offspring from the previous breeding
season) associated with reproductive adult pairs, as described by
Lorenz (1952). The results are summarized in Table 1.
(2) Histograms of neighbour distances. For each of the photo-
graphs used in Table 1, we plotted, for each species, histograms
showing the frequency distribution of neighbour distances. If birds
often ﬂy in discrete dyads one would expect frequency distribu-
tions to exhibit a bimodal character, with the distribution of ﬁrst
neighbour distances being considerably lower than that of the next
six neighbours. As there is no generally accepted formal test of
bimodality, we present the histograms in Appendix Fig. A2 as
qualitative support for the presence of discrete dyads within ﬂocks.
Preferential Association by Species
After controlling for the proportion of rooks within ﬂocks, we
found that a focal bird’s nearest neighbour was signiﬁcantly more
likely to be of the same species (Appendix Table A1).
Occurrence of discrete dyads and triads of jackdaws and rooks in ﬂocks
A 75 10 (27) 6 (24) A 12 8 (67) 0
B 48 12 (50) 2 (13) B 11 4 (36) 0
C 108 12 (22) 5 (14) C 15 4 (27) 3 (20)
D 54 15 (56) 4 (22) E 7 4 (57) 3 (43)
E 19 6 (63) 1 (16) J 43 16 (37) 3 (7)
F 82 17 (41) 7 (26) K 33 14 (42) 0
G 76 12 (32) 2 (8) L 22 12 (55) 0
H 43 10 (47) 1 (7) M 17 10 (59) 6 (35)
I 90 13 (29) 5 (17) N 16 6 (38) 3 (19)
Mean percentages SE 41 5172464146
Numbers in parentheses indicate the percentage of birds of each species ﬂying in
discrete dyads or triads.
J. W. Jolles et al. / Animal Behaviour 85 (2013) 743e750 745
Positional Differences by Species
Rooks made up only 21.8 0.03% of ﬂocks on average, but were
disproportionately likely to be positioned at the front of ﬂocks
(Fig. 1a, Appendix Table A2). The ﬁrst bird at the leading edge was
a rook in 19 out of 44 ﬂocks (¼43.2%), more than twice as often as
expected by chance (binomial test: P¼0.001). Species distributions
within ﬂocks were not signiﬁcantly affected by ﬂock size, month or
time to sunset (Appendix Table A2).
Proximity and Alignment between Neighbours
Neighbours ﬂew more closely together in larger ﬂocks (Fig. 1b,
Appendix Table A3) and in the middle of ﬂocks relative to the front
and back (Fig. 1c, Appendix Table A3). Jackdaw dyads ﬂew sig-
niﬁcantly closer together than rook dyads or mixed dyads (Fig. 2a,
Appendix Table A3), and the directional alignment of same-species
dyads was greater than that of mixed dyads (Fig. 2b, Appendix
Do Birds Fly in Discrete Dyads?
An average of 41 5% of jackdaws (range 22e63%) and 46 4%
(range 37e67%) of rooks in the illustrative selection of photographs
ﬂew in clearly identiﬁable, discrete dyads (Fig. 2c, Table 1). Histo-
grams of neighbour distances commonly showed a bimodal char-
acter with a peak before the average nearest-neighbourdistance for
each species (Figure A2), suggestive of discrete dyads of birds ﬂying
in close proximity.
Contrary to the assumptions of many mathematical models of
single-species aggregations, which treat individuals as equivalent
and interchangeable, our results suggest that the structure of
mixed-species ﬂocks may be critically inﬂuenced by species dif-
ferences and social systems. The larger and socially dominant
rooks were disproportionately likely to be located in the front of
ﬂocks. This effect is unlikely to result from the inﬂuence of par-
ticular individual rooks, as our data set contained photographs of
numerous ﬂocks of differing size, but rather seems to represent
a general property of mixed rookejackdaw ﬂocks. Nor is the
pattern readily explicable by species differences in ﬂight velocity
as rooks tend to be found towards the front of ﬂocks despite
observational evidence suggesting that jackdaws can ﬂy faster
(Coombs 1961). Previous work on ﬁsh schools (Krause et al.
2000), zebra herds (Fischhoff et al. 2007) and small pigeon
ﬂocks (Nagy et al. 2010) suggests that individuals located at the
00 100 200 300
400 500 600
Neighbour distance (jackdaw lengths)
Neighbour distance (jackdaw lengths) Proportion of rooks
Figure 1. (a) Proportion of rooks in the front, middle and back of ﬂocks. The horizontal line indicates the average proportion of rooks across all ﬂocks. (b) Relationship between ﬂock
size and neighbour distances. (c) Distance between neighbours in the front, middle and back of ﬂocks. Bars show means SE. Asterisks indicate signiﬁcance levels between
categories in post hoc analyses: *P<0.05; **P<0.001.
J. W. Jolles et al. / Animal Behaviour 85 (2013) 743e750746
front of groups tend to assume leadership roles, initiating
changes in direction or pace of movement that are followed by
group members. Similarly, rooks may play a dominant role in
inﬂuencing collective movements of mixed-species corvid ﬂocks.
It is possible that rooks’preference for the front of ﬂocks may
simply reﬂect their motivation to reach the roost ﬁrst and obtain
favoured positions (Coombs 1961). If this was the case, one might
expect rooks to move to the front as sunset approaches, but we
found no such effect. Moreover, roosting ﬂocks form spectacular,
swirling displays similar to starling murmurations (King &
Sumpter 2012) before settling, so individuals at the front of
preroosting ﬂocks may not necessarily land ﬁrst at the roost.
Thus, it remains unclear whether rooks derive beneﬁts from
positioning themselves towards the front of ﬂocks, whether jack-
daws preferentially follow rooks or whether species’relative posi-
tions reﬂect aerodynamic considerations Future workincorporating
GPS technology to track ﬂock members (Nagy et al. 2010) could
assist in discriminating between these possibilities.
The general rules of attraction, short-range repulsion and
alignment among neighbours proposed by models of self-
organization provide a valuable framework for understanding
ﬂocking (Bajec & Heppner 2009;Petit & Bon 2010), but our results
indicate that their speciﬁc manifestations may be inﬂuenced by the
characteristics of social systems. Our measurements of neighbour
distances and alignments are somewhat crude and, given the noise
in the data, they are likely to underestimate the true extent of
spatial structure within ﬂocks. Nevertheless, a number of impor-
tant patterns were apparent. First, the extent of attraction and
repulsion may vary depending on the position within a ﬂock, the
size of the ﬂock (see Beauchamp 2012 for similar results in semi-
palmated sandpipers, Calidris pusilla) and the relationships be-
tween group members. Critically, corvids were not evenly
distributed across the ﬂock but typically ﬂew near conspeciﬁcs,
with jackdaws being particularly closely attracted to same-species
neighbours, and birds of both species often appeared to ﬂy in dis-
crete dyads. The occurrence of discrete dyads of birds would not be
Neighbour distance (jackdaw lengths)
Neighbour alignment (degrees)
Mixed Jackdaws Rooks
Figure 2. (a) Distance and (b) alignment between neighbours in jackdaw, rook and mixed dyads. Bars show means SE. Asterisks indicate signiﬁcance levels between categories in
post hoc analyses: **P<0.001. (c) Jackdaws ﬂying in clearly identiﬁable, discrete dyads.
J. W. Jolles et al. / Animal Behaviour 85 (2013) 743e750 747
expected to emerge from morphological or aerodynamic con-
straints alone and is likely to result from social partners ﬂying
together, although further studies with identiﬁable individuals
would be needed to conﬁrm this. Second, the alignment of neigh-
bours was signiﬁcantly higher if they were of the same species,
with jackdaw dyads showing near perfect parallel alignment
(a mean difference of only 3.8
). Both species form lifelong,
monogamous pair bonds characterized by high levels of afﬁliative
behaviour and close proximity (Emery et al. 2007), and our results
suggest the possibility that these relationships are reﬂected in ﬂock
Together, our results suggest that the theoretical convenience of
treating group members as identical and interchangeable does not
adequately reﬂect biological reality in mixed-species ﬂocks. Indeed,
we would argue that this assumption is similarly unlikely to hold in
single-species ﬂocks in which individuals vary and have social
relationships. Differences between individuals can give rise to
leadership roles, which may be particularly pronounced in mixed-
species aggregations in which larger and more dominant species
may commonly take the lead (King et al. 2009). Moreover, studies
of both single-species and mixed-species ﬂocks must consider how
the relationships between individuals may modulate the degree of
attraction, separation and alignment between group members.
Thus, ﬂock structure cannot be fully understood without taking
species’characteristics, their social systems and individuals’re-
lationships into account. Future work incorporating information on
the movements of known individuals will provide further empirical
data that can be integrated into mathematical models to better
understand the inﬂuences of within-group heterogeneity on col-
We thank Neeltje Boogert for her friendship, help and valuable
comments. This work was funded by a British Ecological Society
grant and a BBSRC David Phillips Fellowship to A.T. (BB/H021817/1).
A.J.K was supported by a NERC Fellowship.
Bajec, I. L. & Heppner, F. H. 2009. Organized ﬂight in birds. Animal Behaviour,78,
Ballerini, M., Cabibbo, N., Candelier, R., Cavagna, A., Cisbani, E., Giardina, I.,
Lecomte, V., Orlandi, A., Parisi, G., Procaccini, A., et al. 2008a. Interaction
ruling animal collective behavior depends on topological rather than metric
distance: evidence from a ﬁeld study. Proceedings of the National Academy of
Sciences, U.S.A.,105, 1232e1237.
Ballerini, M., Cabibbo, N., Candelier, R., Cavagna, A., Cisbani, E., Giardina, I.,
Orlandi, A., Parisi, G., Procaccini, A., Viale, M., et al. 2008b. Empirical inves-
tigation of starling ﬂocks: a benchmark study in collective animal behaviour.
Beauchamp, G. 2012. Flock size and density inﬂuence speed of escape waves in
semipalmated sandpipers. Animal Behaviour,83,1125e1129.
Conradt, L. & Roper, T. J. 2005. Consensus decision making in animals. Trends in
Ecology & Evolution,20,449e456.
Coombs, C. J. F. 1961. Rookeries and roosts of the rook and jackdaw in South-West
Cornwall. Bird Study,8,55e70.
Couzin, I. D. & Krause, J. 2003. Self-organization and collective behavior in ver-
tebrates. Advances in the Study of Behavior,32,1e75.
Crawley, M. J. 2002. Statistical Computing. Chichester: J. Wiley.
Emery, N. J., Seed, A. M., von Bayern, A. M. P. & Clayton, N. S. 2007. Cognitive
adaptations of social bonding in birds. Philosophical Transactions of the Royal
Society B,362, 489e505.
Fischhoff, I. R., Sundaresan, S. R., Cordingley, J., Larkin, H. M., Sellier, M. J. &
Rubenstein, D. I. 2007. Social relationships and reproductive state inﬂuence
leadership roles in movements of plains zebra, Equus burchellii.Animal Behav-
Goodale, E. & Beauchamp, G. 2010. The relationship between leadership and
gregariousness in mixed-species bi rd ﬂocks. Journal of Avian Biology,41,
Harcourt, J. L., Ang , T. Z., Sweetman, G., Johnstone, R. A. & Manica, A . 2009.
Social feedback and the emergence of leade rs and followers. Current Biol-
Hemelrijk, C. K. & Hildenbrandt, H. 2011. Some causes of the variable shape of
ﬂocks of Birds. PLoS One,6, e22479.
Jacobs, A., Sueur, C., Deneubourg, J. L. & Petit, O. 2011. Social network inﬂuences
decision making during collective movements in brown lemurs (Eulemur fulvus
fulvus). International Journal of Primatology,32,721e736.
King, A. J. & Sumpter, D. J. T. 2012. Murmurations. Current Biology,22,R112eR114.
King, A. J., Douglas, C. M. S., Huchard, E., Isaac, N. J. B. & Cowlishaw, G. 2008.
Dominance and afﬁliation mediate despotism in a social primate. Current
King, A. J., Johnson, D. D. P. & Van Vugt, M. 2009. The origins and evolution of
leadership. Current Biology,19,R911e916.
Krause, J. 1993. The relationship between foraging and shoal position in a mixed
shoal of roach (Rutilus rutilus) and chub (Leuciscus cephalus): a ﬁeld study.
Krause, J. & Ruxton, G. D. 2002. Living in Groups. Oxford: Oxford University Press.
Krause, J., Hoare, D., Krause, S., Hemelrijk, C. K. & Rubenstein, D. I. 2000.
Leadership in ﬁsh shoals. Fish and Fisheries,1,8
Latta, S. C. & Wunderle, J. M. 1996. Mixed-species ﬂocks in pine forests of Hispa-
niola. The Condor,98, 595e607.
Lockie, J. D. 1956. Winter ﬁghting in feeding ﬂocks of rooks, jackdaws and carrion
crows. Bird Study,3,180e190.
Lorenz, K. 1952. King Solomon’s Ring. London: Methuen.
Nagy, M., Akos, Z., Biro, D. & Vicsek, T. 2010. Hierarchical group dynamics in pi-
geon ﬂocks. Nature,464, 890e893.
Petit, O. & Bon, R. 2010. Decision-making processes: the case of collective move-
ments. Behavioural Processes,84, 635e647.
Selous, E. 19 31. Thought Transference (or What?) in Birds. London: Constable.
Sridhar, H., Beauchamp, G. & Shanker, K. 2009. Why do birds participate in
mixed-species foraging ﬂocks? A large-scale synthesis. Animal Behaviour,78,
Sumpter, D. J. T. 2006. The principles of collective animal behaviour. Philosophical
Transactions of the Royal Society B,361,5e22.
Vicsek, T. & Zafeiris, A. 2012. Collective motion. Physics Reports,517,71e140.
Coefﬁcient estimates in all tables represent the change in the
dependent variable relative to the baseline category and can thus
be interpreted as measures of effect size.
Figure A1. Map of Madingley and surroundings. Photographs were taken within the
large shaded area. To avoid pseudoreplication, photographs taken within a given
evening were shot from different locations within this area. The hatched area shows
the roost, where ﬂocks would combine into a single large ﬂock and spend the night.
J. W. Jolles et al. / Animal Behaviour 85 (2013) 743e750748
70 20 50
0 5 10 15 20 25 0 5 10 15 20 0 5 10 15
10 30 50
Distance in rook len
Distance in jackdaw lengths
30 35 40 5 10 15 20 25 30 35
0 5 10 15 20 25 30 0 5 10 15 0 5 10 15 20 25
0 5 10 0 5 10 15 0 51015
0 5 10 15 20 0 5 10 15 20 25 0 5 10 15
10 0 5 10 0 5 10 15
Figure A2. Histograms of neighbour distances for (a) jackdaws and (b) rooks. Panels show the frequency distribution for the ﬂocks in Table 1. There was considerable variation in
neighbour distances within and between ﬂocks, resulting in part from variation in ﬂock shape and density. Nevertheless, a number of ﬂocks exhibit a binomial character, with the
frequency distribution of ﬁrst neighbours (dark bars) showing a distinct peak. Critically, these peaks are lower than the mean nearest-neighbour distances of 2.4 jackdaw lengths or
3.4 rook lengths, indicating the presence of discrete same-species dyads of birds ﬂying in close proximity to one another.
GLMM on the probability that the nearest neighbour of the focal bird was a jackdaw
Wald statistic (
Proportion of rooks in ﬂock 50.27 1 <0.001
Focal species (jackdaw, rook) 27.78 1 <0.001
Minimal model Effect size SE
Constant 1.46 0.17
Proportion of rooks in ﬂock 4.95 0.70
Jackdaw 0 0
Rook 1.37 0.26
This analysis used data from 454 neighbour dyads in 44 ﬂocks. The binary response
term (1,0) indicated whether the neighbouring bird was a jackdaw. Flock identity
nested in date was ﬁtted as a random term (estimated variance component SE:
GLMM on factors affecting the proportion of rooks among focal birds and their seven
Wald statistic (
Proportion of rooks in ﬂock 41.11 1 <0.001
Flock section (front, middle, back) 26.61 2 <0.001
Month (Oct, Nov, Dec, Jan, Feb) 8.84 4 0.065
Flock size 1.18 1 0.277
Time relative to sunset (min) 1.19 1 0.275
Minimal model Effect size SE
Constant 0.29 0.20
Proportion of rooks in ﬂock 4.72 0.74
Front 0 0
Middle 0.93 0.24
Back 0.78 0.19
This analysis used data from 115 photographs of 44 ﬂocks, with ﬂock identity nested
in date ﬁtted as a random term (estimated variance component SE:
0.201 0.138). Post hoc analyses by exclusion showed that there were signiﬁcantly
more rooks in the front than in the rest of the ﬂock (front >middle:
P<0.001; front >back:
¼11.07, P<0.001; middle ¼back:
LMM on factors affecting the distance between neighbours
Wald statistic (
(jackdaws, rooks, mixed)
48.95 2 <0.001
(front, middle, back)
17.35 2 <0.001
Flock size 6.09 1 0.019
Minimal model Effect size SE
Constant 1.67 0.07
Jackdaws 0 0
Rooks 0.51 0.09
Mixed 0.31 0.07
Front 0 0
Middle 0.13 0.06
Back 0.16 0.07
Flock size 0.001 0.0004
This analysis used data from 454 neighbour dyads in 44 ﬂocks. The response term
was the distance between each of four focal birds per ﬂock section and its nearest
neighbour, measured in jackdaw lengths, and square-root transformed for analysis.
Flock identity nested in date was ﬁtted as a random term (estimated variance
component SE: 0.065 0.024). Post hoc tests by exclusion showed that jackdaw
dyads ﬂew closer together than rook dyads or mixed dyads (jackdaws <rooks:
¼40.65, P<0.001; jackdaws <mixed dyads:
¼1.64, P¼0.203) and dyads in the middle of the ﬂock were
closer than those in the front or back (middle <front:
¼5.83, P¼0.016; mid-
¼22.94, P<0.001; front <back:
¼5.19, P¼0.023; Fig. 1c).
LMM on factors affecting the difference in alignment between neighbours
Wald statistic (
(jackdaws, rooks, mixed)
26.93 2 <0.001
(front, middle, back)
1.05 2 0.592
Flock size 0.05 1 0.821
0.01 1 0.919
Minimal model Effect size SE
Constant 1.09 0.01
Jackdaws 0 0
Rooks 0.01 0.01
Mixed 0.06 0.01
The analysis used data from 454 neighbour dyads in 44 ﬂocks, with ﬂock identity
nested in date ﬁtted as a random term (estimated variance component SE:
0.001 0.000). The response term was normalized for analysis using a BoxeCox
power transformation. Post hoc tests by exclusion showed that same-species
dyads were more closely aligned than mixed dyads (jackdaws <mixed:
¼25.24, P<0.001; rooks <mixed:
¼15.64, P<0.001; jackdaws ¼rooks:
J. W. Jolles et al. / Animal Behaviour 85 (2013) 743e750750