Effects of social environment and personality on communication in ...

<strong>Effects</strong> <strong>of</strong> <strong>social</strong> <strong>environment</strong> <strong>and</strong> <strong>personality</strong> on communication in male
Siamese fighting fish in an artificial network
Giuliano Matessi a, *, Ricardo J. Matos a,1 , Tom M. Peake a,2 , Peter K. McGregor b,3 , Torben Dabelsteen a,1
a
Animal Behaviour Group, Department <strong>of</strong> Biology, University <strong>of</strong> Copenhagen
b
Centre for Applied Zoology, Cornwall College
article info
Article history:
Received 10 February 2009
Initial acceptance 27 March 2009
Final acceptance 30 September 2009
Available online 11 November 2009
MS. number: 09-00090R
Keywords:
aggression
behavioural type
Betta splendens
communication network
<strong>personality</strong>
Siamese fighting fish
temperament
visual signal
Animals <strong>of</strong>ten express consistent individual differences in
behaviour. Part <strong>of</strong> the individual variation in behaviour may be
caused by r<strong>and</strong>om noise around an adaptive average <strong>and</strong> small
variations in state or context, but in a variety <strong>of</strong> animal species
individual behavioural differences are consistent across different
<strong>social</strong> <strong>and</strong> <strong>environment</strong>al contexts <strong>and</strong> independent <strong>of</strong> sex, age <strong>and</strong>
size (Sih et al. 2004; Carere & Eens 2005; Bell 2007; Reale et al.
2007). For example, some individuals may be more aggressive or
bolder than others. Such consistent individual differences are
commonly termed personalities or temperament (Reale et al.
2007). The fitness benefits associated with a specific behaviour in
a particular context (e.g. aggression during a contest) may become
costs in another (aggression during courtship). Therefore, different
* Correspondence <strong>and</strong> present address: G. Matessi, Syvendehusvej 18, 2750
Ballerup, Denmark.
E-mail address: giuliano.matessi@gmail.com (G. Matessi).
1
R. J. Matos <strong>and</strong> T. Dabelsteen are at the Department <strong>of</strong> Biology, University <strong>of</strong>
Copenhagen, Universitetsparken 15, Copenhagen Ø, DK-2100, Denmark.
2
T. M. Peake is now at Student Services, Frenchay Campus, University <strong>of</strong> the West
<strong>of</strong> Engl<strong>and</strong>, Coldharbour Lane, Bristol, BS16 1QY, U.K.
3
P. K. McGregor is at the Centre for Applied Zoology, Cornwall College, Trenance
Gardens, Newquay, Cornwall, TR7 2LZ, U.K.
Animal Behaviour 79 (2010) 43–49
Contents lists available at ScienceDirect
Animal Behaviour
journal homepage: www.elsevier.com/locate/anbehav
Individuals <strong>of</strong> the same species, sex, age <strong>and</strong> size may differ in suites <strong>of</strong> behaviour traits in a consistent
manner across time <strong>and</strong> may thus represent different personalities. In a communication context, the
<strong>personality</strong> <strong>of</strong> an individual may both affect <strong>and</strong> be affected by the behaviour <strong>of</strong> the individuals
surrounding it within a network. We investigated the effects <strong>of</strong> a change <strong>of</strong> local <strong>social</strong> <strong>environment</strong> on
two behavioural types, ‘persistent’ versus ‘sporadic’ signaller, in Siamese fighting fish, Betta splendens.
Males visually interacted for 1 day in a communication network <strong>of</strong> seven fish in tanks arranged in
a hexagonal grid, while we recorded space use <strong>and</strong> signalling data. We then exchanged the positions <strong>of</strong>
two males with different behavioural types <strong>and</strong> observed them interacting the following day. ‘Persistent’
signallers were unaffected by the treatment, while ‘sporadic’ signallers increased the time spent in the
inner front part <strong>of</strong> their tank, from which they could observe but not interact with the neighbours. Social
instability (i.e. number <strong>of</strong> changed neighbours) raised the signalling levels <strong>of</strong> individuals independently
<strong>of</strong> their behavioural types. We discuss the relationship between information gathering in a communication
network <strong>and</strong> network composition in terms <strong>of</strong> behavioural types <strong>of</strong> its members.
Ó 2009 The Association for the Study <strong>of</strong> Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
0003-3472/$38.00 Ó 2009 The Association for the Study <strong>of</strong> Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.anbehav.2009.09.034
personalities may have different context-dependent fitness <strong>and</strong> the
inherent trade-<strong>of</strong>fs may, in general, hinder individuals from
attaining theoretical fitness optima. There is little available information
about the ecological <strong>and</strong> evolutionary implications <strong>of</strong>
personalities, but there are indications that personalities may affect
dispersal, antipredator behaviour, foraging <strong>and</strong> reproductive
strategies <strong>and</strong>, therefore, play a part in population <strong>and</strong> community
ecology (Sih et al. 2004; Dingemanse & Reale 2005; Reale et al.
2007). Personalities have been found in various types <strong>of</strong> behaviours,
mostly having to do with aggressiveness, boldness <strong>and</strong>
reaction to novelty. Some <strong>of</strong> these include exchange <strong>of</strong> signals as
basic components, but there are no direct investigations, to our
knowledge, <strong>of</strong> the implications <strong>of</strong> personalities for communication.
Reale et al. (2007) have suggested five behavioural categories for
personalities: shyness–boldness; exploration–avoidance; activity;
aggressiveness; sociability. Only the last two <strong>of</strong> these require the
use <strong>of</strong> signals at least to some degree, but frequency <strong>of</strong> signal use
<strong>and</strong> reception per se have not been investigated as specific
<strong>personality</strong> traits. Yet the frequency <strong>of</strong> visual threat display has
been used as an index <strong>of</strong> aggressiveness in fish <strong>and</strong> the rate,
frequency or timing <strong>of</strong> specific vocalizations could be used as an
index <strong>of</strong> aggression or eagerness to mate in birds. In theory, both
visual threat display <strong>and</strong> vocalization rate could show consistent

44
individual variation <strong>and</strong> have different costs <strong>and</strong> benefits in
different contexts, for example when the same signals are used as
threats to rivals as opposed to potential mates, or as courtship as
opposed to territorial vocalizations.
Many animals are to some degree <strong>social</strong> <strong>and</strong> even those living in
single individual territories will interact with <strong>and</strong> be affected by
neighbours. There is as yet little information on how the <strong>social</strong>
<strong>environment</strong> affects individual differences in behaviour (van Oers
et al. 2005; Cr<strong>of</strong>t et al. 2006; Magnhagen 2007). For example, the
relationship between exploratory <strong>and</strong> risk-taking behaviours, two
common <strong>personality</strong> axes, seems to vary according to the <strong>social</strong>
context in perch, Perca fluviatilis, great tits, Parus major, <strong>and</strong> ravens,
Corvus corax. In great tits risk-taking behaviour <strong>of</strong> females <strong>and</strong>
‘slow’ males was more affected by the presence <strong>of</strong> a companion
during a foraging task than was that <strong>of</strong> ‘fast’ males (van Oers et al.
2005). Similar effects <strong>of</strong> gender, <strong>social</strong>ity <strong>and</strong> relatedness on novel
object exploration were found in h<strong>and</strong>-raised ravens (Stöwe et al.
2006; Stöwe & Kotrschal 2007). In individual young perch the
correlation between risk taking <strong>and</strong> exploration was dependent on
the personalities <strong>of</strong> the other members <strong>of</strong> the group <strong>of</strong> four in
which they were tested (Magnhagen 2007). In water striders,
Aquarius remigis, group dynamics <strong>and</strong> individual mating activity
were strongly affected by group composition in terms <strong>of</strong> behavioural
types: very aggressive males chased <strong>of</strong>f females <strong>and</strong> thus
lowered group mating frequency (Sih & Watters 2005). Priming by
earlier interactions with conspecifics will also affect the expression
<strong>of</strong> behaviours normally linked with personalities, such as aggression
<strong>and</strong> boldness (Oliveira et al. 2001). The dependence <strong>of</strong> an
individual’s behaviour on what others do is especially obvious in
communication, where the signalling level <strong>of</strong> one individual in an
interaction conditions the signalling level <strong>of</strong> the other (e.g. overlapping
<strong>and</strong> alternating or song type switching <strong>and</strong> matching in
great tits; Peake et al. 2005). In communication networks an individual’s
signalling behaviour will be affected by the individuals it is
directly interacting with, but also by the presence <strong>and</strong> signalling
behaviour <strong>of</strong> byst<strong>and</strong>ers, which could be potential eavesdroppers
(Doutrelant & McGregor 2000; Doutrelant et al. 2001; Dabelsteen
2005; Matos & Schlupp 2005; Peake 2005; Matessi et al. 2008).
Individuals may, for example, vary in their information-gathering
<strong>and</strong> response strategies, <strong>and</strong> this may be expressed in their
eavesdropping behaviour as well as in their better studied exploratory
behaviour (e.g. McGregor & Dabelsteen 1996; Groothuis &
Carere 2005).
Siamese fighting fish, Betta splendens, show consistent individual
differences in aggressiveness, expressed through signalling
by gill cover erection or lateral displays <strong>and</strong> through direct contact
by tail beats <strong>and</strong> bites. These differences also have measurable
consequences on the movement patterns, especially approach <strong>of</strong>
conspecific neighbours. Time spent near a neighbour was consistent
over long periods <strong>of</strong> time <strong>and</strong> predicted the outcome <strong>of</strong> direct
contests in a series <strong>of</strong> laboratory experiments on male–male
interactions (Bronstein 1994). In pilot experiments with artificial
communication networks <strong>of</strong> seven fish we had observed consistent
differences between individuals in signalling <strong>and</strong> movement
patterns which were stable over 3 days <strong>of</strong> testing. Some individuals
seemed to spend most <strong>of</strong> their time patrolling <strong>and</strong> signalling
aggressively along the borders with their three neighbours, while
others alternated intense signalling periods along the borders with
less active periods away from neighbours, but generally not hiding.
All networks contained a mixture <strong>of</strong> the two behavioural types (R. J.
Matos, T. Peake & G. Matessi, unpublished data). The differences in
aggressive signalling may reflect <strong>personality</strong> types, alternative
signalling or fighting tactics, or may be an expression <strong>of</strong> alternative
mating strategies. Since we could not find any published reference
to functional explanations for these differences in behaviour, we
G. Matessi et al. / Animal Behaviour 79 (2010) 43–49
chose to address them as personalities. The different personalities
may be influenced by the <strong>social</strong> <strong>environment</strong>, since fighting fish
males will respond to some degree to the level <strong>of</strong> aggressive signalling
received from neighbours (Bronstein 1994). We quantified
<strong>and</strong> compared these individual differences in fighting fish behaviour
in a network <strong>and</strong> tested the effects <strong>of</strong> <strong>social</strong> <strong>environment</strong> on
individual variation in movement patterns <strong>and</strong> signalling. If
fighting fish vary in <strong>personality</strong> then: (1) individuals within
a <strong>personality</strong> type will resemble each other in space use <strong>and</strong> signalling
more than individuals between <strong>personality</strong> types; (2) these
behaviours will vary little with time; <strong>and</strong> (3) they will not be
strongly affected by the <strong>social</strong> <strong>environment</strong>. We also assessed the
effects <strong>of</strong> <strong>social</strong> instability in a network on the signalling behaviour
<strong>of</strong> the individuals (Sih & Watters 2005; Matessi et al. 2008).
METHODS
Subjects <strong>and</strong> Experiment Set-up
We purchased the 49 subjects used in the experiment at a local
wholesaler in Copenhagen a few weeks before the experiments <strong>and</strong>
housed them in individual aquaria (245 155 mm <strong>and</strong> 250 mm
high filled with water 150 mm deep) under a 12:12 h light:dark
cycle <strong>and</strong> a constant temperature <strong>of</strong> 27 C. All aquaria (housing <strong>and</strong>
experimental) were filled with aged <strong>and</strong> oxygenated tap water. The
fish were fed ad libitum with commercial flake food (VITA). When
not taking part in trials the males were visually isolated from each
other. We measured st<strong>and</strong>ard <strong>and</strong> total length <strong>of</strong> each fish prior to
testing <strong>and</strong> noted its colour category (red or blue).
The experiments took place between 4 March <strong>and</strong> 1 April 2003.
We ran seven trials, each <strong>of</strong> which lasted 3 days with a 1-day break
between trials. During each trial seven fish were chosen at r<strong>and</strong>om
from those not yet tested <strong>and</strong> placed in individual hexagonal
transparent plastic tanks (200 mm on each side <strong>and</strong> 200 mm deep)
arranged in a honeycomb structure within a large circular arena
(Fig. 1). Water levels in the individual tanks <strong>and</strong> in the arena were
kept equal (approximately 15 cm). Each individual tank had an
Shelter
Individual tank
Outer tank
Figure 1. Artificial network <strong>of</strong> fighting fish. The set-up consists <strong>of</strong> seven transparent
hexagonal plastic tanks (200 mm on each side <strong>and</strong> 200 mm deep) arranged in
a honeycomb structure within a large circular outer tank (grey shaded area). Each
hexagonal tank contained an opaque plastic cylinder (dotted circles) as shelter. Within
each <strong>of</strong> the six outer individual tanks we defined three functional zones: border
(striped); inner front (black); back (white). The photo is a top view <strong>of</strong> the set-up
showing what was visible on the screen during behaviour sampling.

opaque cylinder in the middle (100 mm in diameter) with an
entrance facing away from neighbours, which the fish could use for
shelter. We placed the fish in the tanks on the morning <strong>of</strong> the first
day <strong>of</strong> each trial (‘acclimatize’ hereafter) <strong>and</strong> left them to acclimatize
the whole day without visual contact with each other. On
day 2 (‘network’ hereafter), we removed the opaque barriers
separating the fish <strong>and</strong> allowed them to interact visually for 10 h
continuously. At the end <strong>of</strong> the network day we identified the two
fish with different personalities to be manipulated for the current
trial (see below), replaced the opaque barriers <strong>and</strong> exchanged the
positions <strong>of</strong> their tanks. Fish were, therefore, not in visual contact
for the remaining 1 h <strong>of</strong> light or in the dark. On day 3 (‘exchange’
hereafter), we removed the opaque barriers <strong>and</strong> allowed the fish to
interact visually for 10 h continuously. At the end <strong>of</strong> the exchange
day the fish were replaced in their home tanks. Thus at the end <strong>of</strong>
the exchange day each trial consisted <strong>of</strong> two manipulated fish,
which experienced a change in identity <strong>of</strong> two out <strong>of</strong> three
neighbours, <strong>and</strong> four neighbours to the two manipulated fish which
experienced a change in identity (<strong>and</strong> <strong>personality</strong>) <strong>of</strong> one <strong>of</strong> three
neighbours. The central fish experienced no change in neighbour
identity. The water in all the individual tanks was changed on the
day between trials. The fish were fed at the end <strong>of</strong> each day.
Identifying Personalities
In the current experiment we defined the two behavioural types
(personalities) as persistent individuals which spent most <strong>of</strong> the
time at borders <strong>and</strong> interacted with neighbours almost constantly,
<strong>and</strong> sporadic individuals which alternated intense interactions
with neighbours with apparently inactive periods away from
neighbours. We assessed the <strong>personality</strong> <strong>of</strong> each individual
throughout the network day <strong>of</strong> each trial, <strong>and</strong> at the end <strong>of</strong> the day
scored individuals as persistent or sporadic. We then identified two
individuals with different personalities that occupied tanks at
opposite sides <strong>of</strong> the set-up <strong>and</strong>, therefore, had only the central fish
in common as an immediate neighbour. These became the
manipulated individuals in the current trial <strong>and</strong> we exchanged the
positions <strong>of</strong> their tanks at the end <strong>of</strong> the network day (see above).
The <strong>personality</strong> scoring was performed qualitatively, since the
movement <strong>and</strong> signalling behaviours were recorded by computer
(see below) <strong>and</strong> were not available before the end <strong>of</strong> the whole trial.
Each <strong>of</strong> the manipulated fish had two direct neighbours besides the
central fish, so in each trial we could compare the behaviour <strong>of</strong> the
two manipulated fish (persistent <strong>and</strong> sporadic) <strong>and</strong> <strong>of</strong> two categories
<strong>of</strong> neighbours (neighbour to persistent <strong>and</strong> neighbour to
sporadic). We averaged the behaviour <strong>of</strong> the two individuals in each
<strong>of</strong> these categories for analysis. All categories were assigned to
individual fish on the network day <strong>and</strong> individuals maintained their
category assignment on the exchange day.
Data Collection
We sampled movement patterns by computer-controlled automatic
tracking <strong>and</strong> simultaneously scored signalling behaviour
manually for all seven fish on all 3 days using Ethovision v. 3.0
behaviour sampling s<strong>of</strong>tware (Noldus Information Technology,
Wageningen, The Netherl<strong>and</strong>s) on a desktop computer connected
to a video camera placed directly above the set-up. We could also
observe the fish <strong>and</strong> manually score signalling behaviour at a better
resolution on a screen connected in parallel to the camera. Screen
<strong>and</strong> computer were separated from the set-up by a black curtain so
as to not disturb the fish during sampling. The s<strong>of</strong>tware tracked
movement patterns <strong>of</strong> each individual fish over 10 h (from 1 h after
light to 1 h before dark) at a sampling rate <strong>of</strong> five samples per s.
Although we tracked <strong>and</strong> sampled behaviour from all seven fish, we
G. Matessi et al. / Animal Behaviour 79 (2010) 43–49 45
only analysed <strong>and</strong> present here data for the six external fish, thus
excluding the central fish. The six external fish are comparable, all
having three neighbours <strong>and</strong> a clear ‘safe’ zone in the back <strong>of</strong> the
tank, while the central fish had six neighbours <strong>and</strong> no other safe
zone than the central cylinder. The rest <strong>of</strong> the methods <strong>and</strong> the
results will, therefore, apply to the six external fish alone. For
movement-tracking purposes, each tank was divided into four
functional zones (Fig. 1): border (the approximately 7 cm wide onefish-length
b<strong>and</strong> along the edges <strong>of</strong> the tank facing the three
neighbours); inner front (the inner part <strong>of</strong> the half-tank facing the
three neighbours, near the shelter cylinder); back (the half-tank
furthest away from the neighbours); neutral (inside the shelter
cylinder). We sampled the signalling behaviour <strong>of</strong> each fish in turn
using Ethovision’s event recorder. We sampled behaviour only on
the network <strong>and</strong> exchange days, five times a day for 5 min every
second hour, with the first sampling at the start <strong>of</strong> the first hour.
The fish were, therefore, sampled individually in sequence, starting
at the 12-o’clock tank <strong>and</strong> moving clockwise, ending with the
central fish. We recorded opening <strong>and</strong> closing <strong>of</strong> gill covers, tail
beats <strong>and</strong> attempted bites (Matos et al. 2003). We extracted
movement-tracking data summarized in 1 h intervals, which gave
information on total time spent in each zone for each <strong>of</strong> the 10 h <strong>of</strong>
sampling. We extracted duration <strong>and</strong> frequency <strong>of</strong> gill cover
opening <strong>and</strong> frequency <strong>of</strong> tail beats <strong>and</strong> attempted bites for each
behaviour sampling interval (five per day). We also noted the
presence or absence <strong>of</strong> a bubble nest (floating foam-like structures
built by males, for females to lay their eggs). We performed analyses
with Statistica v. 7.0 (Stats<strong>of</strong>t, Tulsa, OK, U.S.A.). All variables
were checked for normality <strong>and</strong> homogeneity <strong>of</strong> variances <strong>and</strong>
transformed where required.
Control Set-up
To control for the possible confounding effect <strong>of</strong> the time that
had elapsed from the network day to the exchange day (Bronstein
1994) on aggressive behaviour <strong>and</strong> activity, we compared our
results with data from a previous experiment, for which the set-up
was similar to our current one until the end <strong>of</strong> the sixth hour <strong>of</strong> the
third day (T. M. Peake, R. J. Matos & G. Matessi, unpublished data),
<strong>and</strong> in which the positions <strong>of</strong> the fish in the set-up were not
modified from day 2 to day 3. The two behavioural types, persistent
<strong>and</strong> sporadic, were not originally identified during this experiment;
therefore we assigned them by looking at the behaviour <strong>of</strong> all fish in
each <strong>of</strong> seven control networks on day 2 (corresponding to the
network day in the manipulated networks) <strong>and</strong> finding movement
<strong>and</strong> signalling patterns that matched those <strong>of</strong> persistent <strong>and</strong>
sporadic fish on the network day <strong>of</strong> the manipulated networks. The
two behavioural types had to be on opposite sides <strong>of</strong> the set-up. We
then analysed the change in their behaviour from day 2 to day 3
(corresponding to the exchange day in manipulated networks). For
this purpose we used only the behaviour in the first 6 h <strong>of</strong> the day,
in which control <strong>and</strong> manipulated networks matched. From the
seventh hour <strong>of</strong> the third day onwards the control <strong>and</strong> manipulated
networks are not comparable, because in the control networks we
placed a physically isolated intruder male in one <strong>of</strong> the hexagonal
tanks, for other purposes than those presented here.
Ethical Note
All fish were housed in individual tanks, visually isolated from
other males. Water was changed regularly <strong>and</strong> ad libitum food was
provided daily. All fish were returned to their home tank after use
<strong>and</strong> kept in our housing facility until their natural death (3–12
months). Although physically isolated from their neighbours, in
theory fish could still harm themselves by attempting to bite the

46
opponent <strong>and</strong> thus hitting the tank wall. Attempted bites were rare,
mostly present right after the opaque barriers were removed. We
inspected the fish after the experiments <strong>and</strong> never found traces <strong>of</strong>
injury.
RESULTS
The composition <strong>of</strong> the networks in terms <strong>of</strong> behavioural types
was variable, with a median <strong>of</strong> two persistent individuals per
network (range 1–4). There seemed to be no strong bias in the
behavioural type <strong>of</strong> the neighbours to the exchange subjects, <strong>and</strong>
most <strong>of</strong> the subjects had one neighbour <strong>of</strong> each type. The seven
persistent subjects had eight persistent <strong>and</strong> six sporadic neighbours,
while the seven sporadic subjects had 11 sporadic <strong>and</strong> three
persistent neighbours. Unfortunately our sample is too small to test
the differences statistically (individuals are not independent within
a network). We did not find any strong pattern in the presence or
absence <strong>of</strong> bubble nests, although at the end <strong>of</strong> the network day no
persistent subject had a bubble nest, while five sporadic subjects
did. At the end <strong>of</strong> the exchange day, five persistent <strong>and</strong> six sporadic
subjects had bubble nests. There was no difference in frequency <strong>of</strong>
colour morph between the subjects (four blue <strong>and</strong> two red
persistent, three blue <strong>and</strong> three red sporadic) <strong>and</strong> no significant
size difference between persistent <strong>and</strong> sporadic subjects (mean -
SE st<strong>and</strong>ard length: persistent: 4.02 0.05 cm; sporadic:
4.0 0.02 cm; paired t test: t5 ¼ 0.35, P ¼ 0.7; total length:
persistent: 8.17 0.21 cm; sporadic: 7.75 0.26 cm; t5 ¼ 2.06,
P ¼ 0.09).
Movement Patterns
We compared use <strong>of</strong> the different zones between <strong>personality</strong>
categories <strong>and</strong> between days averaged over the first 6 h <strong>of</strong> observation,
in which manipulated <strong>and</strong> control networks are comparable.
In the manipulated set-up we found highly significant effects
<strong>of</strong> treatment day, <strong>personality</strong> category <strong>and</strong> their interaction on the
time spent by an individual at the border (repeated measures
ANOVA: day: F2,12 ¼ 24.498, P ¼ 0.00006; <strong>personality</strong>:
F3,18 ¼ 10.797, P ¼ 0.0003; interaction: F6,36 ¼ 4.812, P ¼ 0.001;
Fig. 2a). The four <strong>personality</strong> categories were not significantly
different on the acclimatize day (Tukey HSD: all P > 0.99). On the
network day persistent <strong>and</strong> sporadic were significantly different, as
expected from the definition <strong>of</strong> the two <strong>personality</strong> types (Tukey
HSD: P ¼ 0.0001). After the exchange, sporadic increased slightly
the time spent at the border compared to the previous day, but the
change in time at the border (exchange day minus network day)
was not significantly different from zero (bootstrap r<strong>and</strong>omization
test: P ¼ 0.14), possibly caused by the increase in variation (Fig. 2a).
Sporadic was still significantly different from persistent (Tukey
HSD: P ¼ 0.0003). We found weakly significant effects <strong>of</strong> <strong>personality</strong>
category <strong>and</strong> <strong>of</strong> the interaction term on the time in the inner
front zone (day: F2,12 ¼ 1.553, P ¼ 0.25; <strong>personality</strong>: F3,18 ¼ 3.351,
P ¼ 0.04; interaction: F6,36 ¼ 2.764, P ¼ 0.03; Fig. 2b). Sporadic
seemed to spend more time than persistent in the inner front zone
on the exchange day. There were no significant differences between
the <strong>personality</strong> categories within either the acclimatize or network
day (Tukey HSD: all P > 0.24). Sporadic spent significantly more
time than persistent in the inner front zone on the exchange day
(Tukey HSD: P ¼ 0.001). The change in use <strong>of</strong> the inner front zone
(Fig. 2b) by sporadic individuals was significantly different from
zero (bootstrap r<strong>and</strong>omization test: P < 0.05). Similar results were
obtained when looking at the movement patterns over the whole
day (10 h <strong>of</strong> observation), except that none <strong>of</strong> the effects was
significant for the use <strong>of</strong> the inner front zone (repeated measures
ANOVA: all F < 1.776, all P > 0.19).
G. Matessi et al. / Animal Behaviour 79 (2010) 43–49
Proportion <strong>of</strong> time at border
Proportion <strong>of</strong> time at inner front
1.2
1
0.8
0.6
0.4
0.2
0
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
(a)
(b)
Acclimatize
Acclimatize
In the control set-up the movement patterns <strong>of</strong> the fish were
affected significantly also by their behavioural type <strong>and</strong> by the 3
days <strong>of</strong> testing (repeated measures ANOVA: time spent by an
individual at the border: day: F2,12 ¼ 22.6, P ¼ 0.0001; <strong>personality</strong>:
F3,18 ¼ 6.5, P ¼ 0.004; interaction: F6,36 ¼ 2.8, P ¼ 0.02; time in the
inner front zone: day: F2,12 ¼ 24.533, P < 0.0001; <strong>personality</strong>:
F3,18 ¼ 4.52, P ¼ 0.016; interaction: F6,36 ¼ 2.738, P ¼ 0.027). On the
other h<strong>and</strong>, in this set-up the change in use <strong>of</strong> either the border
zone or the inner front zone from day 2 to day 3 by sporadic
individuals, or any <strong>of</strong> the other three behavioural types, did not
differ significantly from zero (bootstrap r<strong>and</strong>omization test:
P > 0.49 for both variables).
The movement patterns <strong>of</strong> persistent individuals on the
network day were correlated significantly <strong>and</strong> positively with those
on the exchange day (time at border zone: r ¼ 0.77, P ¼ 0.04; time
in inner front zone: r ¼ 0.79, P ¼ 0.03), but those <strong>of</strong> sporadic individuals
were not (time at border zone: r ¼ 0.52, P ¼ 0.23; time in
inner front zone: r ¼ 0.65, P ¼ 0.11; df ¼ 5 in all correlations).
Signalling <strong>and</strong> Aggressive Behaviour
Network
Network
Day
Exchange
Exchange
Figure 2. Space use by the two experimental fish (persistent ¼ solid squares, sporadic
¼ solid circles) <strong>and</strong> their neighbours (neighbour to persistent ¼ empty squares,
neighbours to sporadic ¼ empty circles). The graphs represent mean SE <strong>of</strong>: (a)
proportion <strong>of</strong> time spent at the border zone, <strong>and</strong> (b) proportion <strong>of</strong> time spent at the
inner front zone <strong>of</strong> the tank, averaged over the first 6 h <strong>of</strong> recording.
We compared gill cover displays <strong>and</strong> tail beat use among
<strong>personality</strong> categories <strong>and</strong> between the network <strong>and</strong> exchange
days for the first three observation periods, corresponding to the
first 6 h <strong>of</strong> each trial. We found significant effects <strong>of</strong> <strong>personality</strong>
category on both the duration <strong>and</strong> the frequency <strong>of</strong> gill cover use
(repeated measures ANOVA: gill cover frequency: day: F1,6 ¼ 3.071,
P ¼ 0.13; <strong>personality</strong>: F3,18 ¼ 22.86, P ¼ 0.000002; interaction:
F3,18 ¼ 0.942, P ¼ 0.44; gill cover duration: day: F1,6 ¼ 0.232,

P ¼ 0.65; <strong>personality</strong>: F3,18 ¼ 7.507, P ¼ 0.002; interaction:
F3,18 ¼ 0.835, P ¼ 0.5; Fig. 3a, b). On the network day, persistent <strong>and</strong>
sporadic were significantly different in the frequency <strong>of</strong> gill cover
display, with persistent signalling more <strong>of</strong>ten than sporadic
(Tukey’s HSD: P ¼ 0.003), but not in duration <strong>of</strong> gill cover display
(P ¼ 0.1). After the exchange, sporadic was significantly different
from persistent in both signalling variables (Tukey’s HSD: gill cover
display frequency: P ¼ 0.001; gill cover display duration: P ¼ 0.02).
We found no significant difference between <strong>personality</strong> categories
or between days in the frequency <strong>of</strong> use <strong>of</strong> tail beats (day:
F1,6 ¼ 0.358, P ¼ 0.6; <strong>personality</strong>: F3,18 ¼ 1.898, P ¼ 0.2; interaction:
F3,18 ¼ 1.763, P ¼ 0.2; Fig. 3c). Attempted bites occurred too rarely to
be analysed statistically (30 in total) <strong>and</strong> did not show clear
Frequency <strong>of</strong> gill cover display
Duration <strong>of</strong> gill cover displays (s)
Frequency <strong>of</strong> tail beats
30
25
20
15
10
5
0
140
120
100
80
60
40
20
25
20
15
10
5
0
-5
0
(a)
(b)
(c)
Network
Network
Network
Day
Exchange
Exchange
Exchange
Figure 3. Signalling behaviour by the two experimental fish (persistent ¼ solid
squares, sporadic ¼ solid circles) <strong>and</strong> their neighbours (neighbour to persistent
¼ empty squares, neighbours to sporadic ¼ empty circles) across the network <strong>and</strong>
exchange days <strong>of</strong> the experiment. The graphs represent mean SE <strong>of</strong>: (a) frequency <strong>of</strong>
gill cover displays, (b) duration <strong>of</strong> gill cover displays (s), <strong>and</strong> (c) frequency <strong>of</strong> tail beats,
averaged over the first three behaviour sampling intervals.
G. Matessi et al. / Animal Behaviour 79 (2010) 43–49 47
occurrence patterns. Similar results were obtained by analysing the
signalling behaviour over the whole trial (five observation periods).
As far as signalling behaviour in the control set-up is concerned,
day <strong>and</strong> <strong>personality</strong> had a significant effect only on the number <strong>of</strong>
tail beats used (day: F1,6 ¼ 14.764, P ¼ 0.008; <strong>personality</strong>:
F3,18 ¼ 3.325, P ¼ 0.04; interaction: F3,18 ¼ 1.125, P ¼ 0.36).
<strong>Effects</strong> <strong>of</strong> Social Instability
The behaviour <strong>of</strong> the six fish in each network can be analysed by
grouping the fish according to the number <strong>of</strong> ‘new’ neighbours they
were confronted with on the exchange day, which can be considered
a measure <strong>of</strong> <strong>social</strong> instability. Each neighbour to the
exchanged fish was confronted with one new neighbour on the
exchange day, while the exchanged fish were faced with two new
neighbours. We calculated average values <strong>of</strong> movement pattern
<strong>and</strong> aggressive signalling behaviour for each <strong>of</strong> these two new
categories (one versus two neighbours changed) <strong>and</strong> for each <strong>of</strong> the
seven networks for the first three observation periods. We then
compared the two categories on the network <strong>and</strong> exchange days.
We found a weakly significant effect <strong>of</strong> the number <strong>of</strong> new neighbours
on the frequency <strong>of</strong> gill cover display (repeated measures
ANOVA: day: F1,6 ¼ 3.072, P ¼ 0.13; number <strong>of</strong> changed neighbours:
F1,6 ¼ 6.007, P ¼ 0.05; interaction: F1,6 ¼ 1.031, P ¼ 0.35;). The individuals
with two changed neighbours used gill cover displays more
frequently on the exchange day than those with only one (P ¼ 0.03).
We calculated the same variables for the control set-ups, where no
change <strong>of</strong> neighbours took place, by pooling the data for sporadic
<strong>and</strong> persistent (equivalent to the two changed neighbours group)
<strong>and</strong> for the neighbours to sporadic <strong>and</strong> persistent (equivalent to the
one changed neighbour group). We found no significant differences
between these groups in any <strong>of</strong> the variables tested.
DISCUSSION
Individual fighting fish differed in behaviour when allowed to
interact in an artificial network, expressing aggressive signalling in
either a persistent or a sporadic manner. Generally, behavioural
types were reasonably stable across changing <strong>social</strong> <strong>environment</strong>s,
since the behaviour <strong>of</strong> persistent individuals was significantly
positively correlated before <strong>and</strong> after the exchange, <strong>and</strong> that <strong>of</strong>
sporadic individuals was also positively (but not significantly)
correlated before <strong>and</strong> after the exchange. However, the <strong>social</strong>
<strong>environment</strong> seemed to have different effects on these two
behavioural types. Persistent individuals maintained their behaviour
patterns after changing position in the network. Sporadic
individuals had a tendency to become more similar to persistent
individuals in some respects, signalling more <strong>and</strong> spending relatively
more time near the border, but did not change behaviour
significantly. The most striking difference was that sporadic individuals
seemed to spend more time in the inner front zones, an area
<strong>of</strong> their territory from which they could potentially collect information
from neighbours without interacting with them. Fish
whose position in the network stayed the same did not show
a corresponding change in behaviour, indicating that it is the
change itself that affects behavioural types <strong>and</strong> not a time effect
(Bronstein 1994). The neighbours <strong>of</strong> exchanged individuals did not
modify their behaviour, although there seemed to be a potentially
interesting slight trend for them to ‘switch’ signalling patterns, that
is, neighbours to persistent became more like neighbours to
sporadic <strong>and</strong> vice versa, possibly responding to the change in signalling
they received. Aggressive signalling was more frequent with
more direct neighbours changing identity. Social instability,
therefore, significantly affected the signalling behaviour <strong>of</strong> individuals
independently <strong>of</strong> their behavioural type.

48
The results <strong>of</strong> our experiment can be discussed under three
main headings: (1) information gathering in a network seems to be
a potentially important component <strong>of</strong> individual behavioural types;
(2) behavioural types may have different degrees <strong>of</strong> flexibility; (3)
<strong>social</strong> instability affects signalling behaviour within a network.
Sporadic signallers on average seemed to increase the time spent in
the inner front zone <strong>of</strong> their tank after two <strong>of</strong> their neighbours
changed identity (i.e. after the exchange). The inner front zone is
a position from which it is possible to observe neighbours <strong>and</strong>
especially their interactions, without taking part in them <strong>and</strong> with
a potentially lower risk <strong>of</strong> being attacked. It could therefore
represent a good information-gathering <strong>and</strong> eventually eavesdropping
position within the network (Oliveira et al. 1998;
McGregor et al. 2001; Dabelsteen 2005; Peake 2005; Mathevon
et al. 2005; Matos & Schlupp 2005). The instability created by the
exchange, or in any case the maintenance <strong>of</strong> instability beyond that
<strong>of</strong> the network day (where neighbours are seen for the first time),
may have required the collection <strong>of</strong> new information, <strong>and</strong> the two
behavioural types may gather information in different ways, for
example persistent signallers may preferentially gather information
by direct interactions, while sporadic signallers may to a higher
degree prefer to gather further indirect information through
eavesdropping (Oliveira et al. 1998). Of course these suggestions
would require further, specific testing to be confirmed.
The movement pattern <strong>and</strong> signalling behaviour <strong>of</strong> persistent
signallers were less affected by the changes in <strong>social</strong> <strong>environment</strong>
than those <strong>of</strong> sporadic signallers, which seems to indicate that one
behavioural type may be somewhat more stable than the other (e.g.
Ruiz-Gomez et al. 2008). The individual variability in response to the
manipulation seemed to differ between the two behavioural types,
as judged qualitatively by the spread <strong>of</strong> the st<strong>and</strong>ard errors (Fig. 3).
Therefore, it is possible that ‘sporadic’ signallers were in fact a more
heterogeneous group <strong>of</strong> individuals than ‘persistent’ signallers.
Differences in flexibility between personalities were found among
great tits, where fast individuals were less affected by the presence
<strong>of</strong> a conspecific during a foraging task than slow individuals, possibly
because fast individuals develop routine behaviour quickly <strong>and</strong> are
less sensitive to external signals (van Oers et al. 2005). Persistent
fighting fish may do the same <strong>and</strong> quickly lock into a pattern, which
may even be reinforced by the reaction <strong>of</strong> their neighbouring
opponents. Our results are consistent in part with previous findings
on fighting fish, in which the aggression levels <strong>of</strong> individuals classified
as highly aggressive did not change with increased duration <strong>of</strong>
exposure to neighbours, whereas those <strong>of</strong> less aggressive individuals
did (Bronstein 1994). This effect was not only due to the duration <strong>of</strong>
exposure in our case, since the fish in the control networks did not
show a corresponding effect, but was at least in part caused by
changes in <strong>social</strong> <strong>environment</strong>.
Social instability, in general, may have an effect on some aspects
<strong>of</strong> individual signalling. Changing the identity <strong>of</strong> more neighbours
produced higher average signalling levels, although this effect was
limited to only one <strong>of</strong> the signalling behaviours measured, <strong>and</strong> only
in the first part <strong>of</strong> the day. The effect <strong>of</strong> the instability cannot be
caused by the need to assess new individuals by interacting with
them, because the neighbours are also ‘new’ <strong>and</strong> need to be
assessed on the morning <strong>of</strong> the network day. The individuals must
somehow recognize the neighbours as different from those
encountered the day before to explain their different responses.
The effect could be the product <strong>of</strong> a habituation – relapse type <strong>of</strong>
response, but with habituation persisting overnight. This seems
unlikely, <strong>and</strong> would in any case require individual discrimination
<strong>and</strong> an additive effect <strong>of</strong> number <strong>of</strong> different individuals encountered
on signalling levels. Increased aggressive signalling by
priming (Matos et al. 2003) would also require some level <strong>of</strong>
additive effects <strong>of</strong> number <strong>of</strong> different individuals encountered. The
G. Matessi et al. / Animal Behaviour 79 (2010) 43–49
responses <strong>of</strong> fighting fish have some similarities with the responses
<strong>of</strong> songbirds to some types <strong>of</strong> song playback experiments testing for
neighbour–stranger discrimination. Playback to territorial songbirds
<strong>of</strong> songs from individuals that are not neighbours (‘strangers’)
commonly elicit stronger responses irrespective <strong>of</strong> loudspeaker
location. However, the response elicited by playback <strong>of</strong> neighbours
depends strongly on location: relatively little response when the
location is close to the usual territory boundary, but very strong
responses when playback is from the opposite side <strong>of</strong> the territory
(an unusual location). The phenomenon was originally described
for the white-throated sparrow, Zonotrichia albicollis (Falls & Brooks
1975), but since has been reported for more than 20 passerine
species (reviewed in Stoddard 1996), <strong>and</strong> later found in numerous
other animal groups. The phenomenon is normally interpreted as
a ‘dear enemy’ effect (Fisher 1954), where a stable neighbour is not
a threat to the territory owner, but an expansionist or otherwise
unstable neighbour suddenly requires attention (e.g. Godard 1993).
This difference in response could be interpreted in the same
information-gathering role as suggested for fighting fish displays in
our study. It would be interesting to relate such neighbour–stranger
playbacks to individual differences (potential <strong>personality</strong> types) in
patterns <strong>of</strong> territorial behaviour. We should underline that our
results on <strong>social</strong> instability are based on a small sample size <strong>and</strong> <strong>of</strong>
limited statistical significance, so they are by no means conclusive.
The mechanism by which signalling levels are modified by <strong>social</strong>
instability is also unclear <strong>and</strong> requires further investigation.
Personalitycan influence the broader <strong>social</strong> network <strong>of</strong> individuals
(Matessi et al. 2008) <strong>and</strong> <strong>personality</strong> composition may affect group
structure. Groups <strong>of</strong> bold <strong>and</strong> shy sticklebacks, Gasterosteus aculeatus,
differed in the number, distribution <strong>and</strong> type <strong>of</strong> pairwise interactions,
mediated by the influence <strong>of</strong> <strong>personality</strong> on movement patterns
during interactions <strong>and</strong> on position preferences in a shoal (Ward et al.
2004; Pike et al. 2008). We did not find any consistent pattern in
either network composition or behavioural types <strong>of</strong> neighbours to the
subjects, but more specific testing is required. Our results seem to
suggest that individual preferences for information-spreading <strong>and</strong>
gathering strategies may underlie the effects <strong>of</strong> <strong>personality</strong> on <strong>social</strong>
network structure in some cases. Looking at personalities in an
integrated <strong>social</strong> <strong>and</strong> communication network context may help
clarify complex <strong>social</strong> dynamics (Matessi et al. 2008).
Our experiments are, to the best <strong>of</strong> our knowledge, the first
to suggest a relationship between <strong>personality</strong>, signalling <strong>and</strong>
communication network structure, <strong>and</strong> our conclusions must,
therefore, be cautious <strong>and</strong> preliminary. Caution is especially suggested
by the loss <strong>of</strong> statistical significance in some results when
we analysed the behaviour <strong>of</strong> the fish over the whole day. We are
also not yet sure whether the two behavioural types we identified
in fighting fish can be clearly defined as ‘personalities’, <strong>and</strong> how
consistent they are across time <strong>and</strong> context, as we could not find
any clear previous reference to them in the literature. One <strong>of</strong> the
two may indeed be more heterogeneous or flexible than the other.
The two behavioural types may indeed represent different breeding
strategies rather than personalities, where sporadic individuals
invest more in nest construction <strong>and</strong> maintenance (bubble nests
are <strong>of</strong>ten placed away from borders with neighbouring territories,
personal observation), <strong>and</strong> persistent individuals invest in defence
<strong>and</strong> mate attraction. The difference between personalities <strong>and</strong>
strategies is also somewhat unclear in the literature <strong>and</strong> the two
concepts may well be closely related (Sih et al. 2004; Bell 2007;
Reale et al. 2007). We will need further tests <strong>of</strong> repeatability <strong>and</strong>
context dependence <strong>of</strong> ‘personalities’ in signalling <strong>and</strong> communication
behaviour in fighting fish <strong>and</strong> other species, but our results
seem to indicate that individual variation in signalling <strong>and</strong> information
gathering may influence information flow <strong>and</strong> the structure
<strong>of</strong> communication networks.

Acknowledgments
We thank Denise Pope, Andrew Terry <strong>and</strong> other students <strong>and</strong>
colleagues <strong>of</strong> the Animal Behaviour Group at the Department <strong>of</strong>
Biology for their support <strong>and</strong> for the many constructive discussions
on the project. We thank Anne Marie Wolthers Hansen for
managing housing <strong>and</strong> feeding <strong>of</strong> the fish <strong>and</strong> John Bruun Andresen
for building the set-up. The project was funded by Framework
Grant no. 21-04-0403 to T.D. <strong>and</strong> by SNF grant no. 9801928 to P.K.M.
G.M. was supported by an EU Marie Curie Individual Fellowship
(HPMF-CT-2001-01474).
References
Bell, A. M. 2007. Future directions in behavioural syndromes research. Proceedings
<strong>of</strong> the Royal Society B, 274, 755–761.
Bronstein, P. M. 1994. On the predictability, sensitization, <strong>and</strong> habituation <strong>of</strong> aggression
in male bettas (Betta splendens). Journal <strong>of</strong> Comparative Psychology, 108, 45–57.
Carere, C. & Eens, M. 2005. Unravelling animal personalities: how <strong>and</strong> why individuals
consistently differ. Behaviour, 142, 1149–1157.
Cr<strong>of</strong>t, D. P., James, R., Thomas, P. O. R., Hathaway, C., Mawdsley, D., Lal<strong>and</strong>, K. N.
& Krause, J. 2006. Social structure <strong>and</strong> co-operative interactions in a wild
population <strong>of</strong> guppies (Poecilia reticulata). Behavioral Ecology <strong>and</strong> Sociobiology,
59, 644–650.
Dabelsteen, T. 2005. Public, private or anonymous? Facilitating <strong>and</strong> countering
eavesdropping. In: Animal Communication Networks (Ed. by P. K. McGregor),
pp. 38–62. Cambridge: Cambridge University Press.
Dingemanse, N. J. & Reale, D. 2005. Natural selection <strong>and</strong> animal <strong>personality</strong>.
Behaviour, 142, 1159–1184.
Doutrelant, C. & McGregor, P. K. 2000. Eavesdropping <strong>and</strong> mate choice in female
fighting fish. Behaviour, 137, 1655–1669.
Doutrelant, C., McGregor, P. K. & Oliveira, R. F. 2001. The effect <strong>of</strong> an audience on
intrasexual communication in male Siamese fighting fish, Betta splendens.
Behavioral Ecology, 12, 283–286.
Falls, J. B. & Brooks, J. R. 1975. Individual recognition by song in white-throated
sparrows. II. <strong>Effects</strong> <strong>of</strong> location. Canadian Journal <strong>of</strong> Zoology, 53, 412–420.
Fisher, R. A. 1954. Evolution <strong>and</strong> bird <strong>social</strong>ity. In: Evolution As a Process (Ed. by
J. Huxley, A. C. Hardy & E. B. Ford), pp. 71–83. London: Allen & Unwin.
Godard, R. 1993. Red-eyed vireos have difficulty recognizing individual neighbors’
songs. Auk, 110, 857–862.
Groothuis, T. G. G. & Carere, C. 2005. Avian personalities: characterization <strong>and</strong>
epigenesis. Neuroscience <strong>and</strong> Biobehavioral Reviews, 29, 137–150.
McGregor, P. K. & Dabelsteen, T. 1996. Communication networks. In: Ecology <strong>and</strong>
Evolution <strong>of</strong> Acoustic Communication in Birds (Ed. by D. E. Kroodsma &
E. H. Miller), pp. 409–425. Ithaca, New York: Cornell University Press.
McGregor, P. K., Peake, T. M. & Lampe, H. M. 2001. Fighting fish Betta splendens
extract relative information from apparent interactions: what happens when
what you see is not what you get. Animal Behaviour, 62, 1059–1065.
Magnhagen, C. 2007. Social influence on the correlation between behaviours in
young-<strong>of</strong>-the-year perch. Behavioral Ecology <strong>and</strong> Sociobiology, 61, 525–531.
G. Matessi et al. / Animal Behaviour 79 (2010) 43–49 49
Matessi, G., Matos, R. J. & Dabelsteen, T. 2008. Communication in <strong>social</strong> networks
<strong>of</strong> territorial animals: networking at different levels in birds <strong>and</strong> other systems.
In: Sociobiology <strong>of</strong> Communication (Ed. by P. D’Ettorre & D. P. Hughes), pp. 33–54.
Oxford: Oxford University Press.
Mathevon, N., Dabelsteen, T. & Blumenrath, S. H. 2005. Are high perches in the
blackcap Sylvia atricapilla song or listening posts? A sound transmission study.
Journal <strong>of</strong> the Acoustical Society <strong>of</strong> America, 117, 442–449.
Matos, R. J. & Schlupp, I. 2005. Performing in front <strong>of</strong> an audience: signallers <strong>and</strong>
the <strong>social</strong> <strong>environment</strong>. In: Animal Communication Networks (Ed. by
P. K. McGregor), pp. 63–83. Cambridge: Cambridge University Press.
Matos, R. J., Peake, T. M. & McGregor, P. K. 2003. Timing <strong>of</strong> presentation <strong>of</strong> an
audience: aggressive priming <strong>and</strong> audience effects in male displays <strong>of</strong> Siamese
fighting fish (Betta splendens). Behavioral Processes, 63, 53–61.
van Oers, K., Klunder, M. & Drent, P. J. 2005. Context dependence <strong>of</strong> personalities:
risk-taking behavior in a <strong>social</strong> <strong>and</strong> a non<strong>social</strong> situation. Behavioral Ecology, 16,
716–723.
Oliveira, R. F., McGregor, P. K. & Latruffe, C. 1998. Know thine enemy: fighting fish
gather information from observing conspecific interactions. Proceedings <strong>of</strong> the
Royal Society B, 265, 1045–1049.
Oliveira, R. F., Lopes, M., Carneiro, L. A. & Canario, A. V. M. 2001. Watching fights
raises fish hormone levels: cichlid fish wrestling for dominance induce an
<strong>and</strong>rogen surge in male spectators. Nature, 409, 475.
Peake, T. M. 2005. Eavesdropping in communication networks. In: Animal
Communication Networks (Ed. by P. K. McGregor), pp. 13–27. Cambridge: Cambridge
University Press.
Peake, T. M., Matessi, G., McGregor, P. K. & Dabelsteen, T. 2005. Song type
matching, song type switching <strong>and</strong> eavesdropping in male great tits. Animal
Behaviour, 69, 1063–1068.
Pike, T. W., Samanta, M., Lindstrom, J. & Royle, N. J. 2008. Behavioural phenotype
affects <strong>social</strong> interactions in an animal network. Proceedings <strong>of</strong> the Royal Society
B, 275, 2515–2520.
Reale, D., Reader, S. M., Sol, D., McDougall, P. T. & Dingemanse, N. J. 2007. Integrating
animal temperament within ecology <strong>and</strong> evolution. Biological Reviews,
82, 291–318.
Ruiz-Gomez, M. D., Kittilsen, S., Höglund, E., Huntingford, F. A., Sørensen, C.,
Pottinger, T. G., Bakken, M., Winberg, S., Korzan, W. J. & Øverli, Ø 2008.
Behavioral plasticity in rainbow trout (Oncorhynchus mykiss) with divergent
coping styles: when doves become hawks. Hormones <strong>and</strong> Behavior, 54,
534–538.
Sih, A. & Watters, J. V. 2005. The mix matters: behavioural types <strong>and</strong> group
dynamics in water striders. Behaviour, 142, 1417–1431.
Sih, A., Bell, A. M., Johnson, J. C. & Ziemba, R. E. 2004. Behavioral syndromes: an
integrative overview. Quarterly Review <strong>of</strong> Biology, 79, 241–277.
Stoddard, P. K. 1996. Vocal recognition <strong>of</strong> neighbors by territorial passerines. In:
Ecology <strong>and</strong> Evolution <strong>of</strong> Acoustic Communication in Birds (Ed. by D. E. Kroodsma
& E. H. Miller), pp. 356–374. Ithaca, New York: Cornell University Press.
Stöwe, M. & Kotrschal, K. 2007. Behavioural phenotypes may determine whether
<strong>social</strong> context facilitates or delays nove object exploration in ravens (Corvus
corax). Journal <strong>of</strong> Ornithology, 148 (Supplement. 2), S179–S184.
Stöwe, M., Bugnyar, T., Loretto, M. C., Schloegl, C., Range, F. & Kotrschal, K. 2006.
Novel object exploration in ravens (Corvus corax): effect <strong>of</strong> <strong>social</strong> relationship.
Behavioral Processes, 73, 68–74.
Ward, A. J. W., Thomas, P., Hart, P. J. B. & Krause, J. 2004. Correlates <strong>of</strong> boldness in
three-spined sticklebacks (Gasterosteus aculeatus). Behavioral Ecology <strong>and</strong>
Sociobiology, 55, 561–568.