Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Animal Behaviour 83 (2012) 783e791

Contents lists available at SciVerse ScienceDirect

Animal Behaviour
journal homepage: www.elsevier.com/locate/anbehav

Changes in the visual environment affect colour signal brightness and


shoaling behaviour in a freshwater fish
Jennifer L. Kelley*, Bree Phillips 1, Gabrielle H. Cummins 2, Julia Shand 3
Centre for Evolutionary Biology, Experimental & Regenerative Neurosciences, School of Animal Biology (M092), University of Western Australia, Perth, Australia

a r t i c l e i n f o
Aquatic organisms are exposed to highly variable light environments, which can affect the efficacy of colour
Article history: patterns that are used for communication or camouflage. Specifically, dissolved organic matter that is
Received 25 August 2011 common in turbid freshwater habitats tends to absorb short wavelength light causing a shift towards
Initial acceptance 4 November 2011 environments that are rich in long wavelengths (orange/red). We investigated how changes in the intensity
Final acceptance 13 December 2011 and wavelength of light affect colour pattern expression and shoaling behaviour in a colourful freshwater
Available online 26 January 2012 fish, the western rainbowfish, Melanotaenia australis. We used light filters to simulate environments rich
MS. number: 11-00676R in organic matter (yellow filters: long wavelength dominated, reduced luminance), habitats with
full-spectrum lighting but with reduced luminance (neutral-density filters) and a control in which no
Keywords: changes in the light environment occurred (no filters). We measured changes in the area and brightness of
adaptive coloration
colour patterns using digital photography and spectrometry and we evaluated the effect of lighting on fish
colour vision
social (shoaling) behaviour. Rainbowfish in the dissolved organic matter treatment showed an increase in
crypsis
Melanotaenia australis
the area and brightness of their colour patterns and individuals shoaled further apart than those in the
rainbowfish control group. The increased brightness of red colours in environments rich in organic matter could act to
shoaling enhance colour pattern conspicuousness, allowing individuals to maintain communication in altered
signalling visual environments. However, an understanding of the species’ visual system is required to determine
turbidity levels of contrast of the colour patterns with respect to variable background environments.
visual ecology Ó 2012 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Animal communication systems are a result of the combined Seehausen et al. 1997; Boughman 2001; Gomez & Thery 2004;
evolution of signal design and sensory perception (Endler 1992). Leal & Fleishman 2004; Cummings 2007; Stuart-Fox et al. 2007;
The environment has an important effect on this evolutionary Morrongiello et al. 2010). However, signals that have evolved for
outcome because the efficacy of a signal depends on the trans- maximum efficacy in one environment may perform poorly in
mission properties of the medium (e.g. air/water) and the sensory another (Leal & Fleishman 2004), which should result in selection
background (e.g. noise, visual complexity) against which the signal for plasticity in signal production. While it is often assumed that
is detected (Endler 1990). Selection must favour the signalling the relationship between signal variation and environmental
system that is most effective in a particular environment, which, in heterogeneity is a result of evolutionary adaptation, few
the case of sexually selected signals, can promote the processes of researchers have considered contextual plasticity in colour signal
reproductive isolation and speciation (Endler 1992; Endler & Basolo expression (Ord et al. 2010).
1998; Boughman 2002). Changes in the environmental conditions encountered during
Numerous studies have demonstrated that the colour patterns signalling often cause animals to alter the intensity and/or duration
of a species or population are often correlated with the composition of their signalling behaviour. In visual communication systems,
or luminance of light in the local habitat (Reimchen 1989; males often perform their courtship displays at the time and/or
location at which their colour patterns will be perceived as most
conspicuous against the background environment (Endler 1987,
* Correspondence: J. L. Kelley, Centre for Evolutionary Biology, Experimental &
1991; Heindl & Winkler 2003; Schultz et al. 2008). In acoustic
Regenerative Neurosciences, School of Animal Biology (M092), University of
Western Australia, Crawley 6009, Perth, WA, Australia. systems, birds increase their dominant song frequency in noisy
E-mail address: jennifer.kelley@uwa.edu.au (J. L. Kelley). urban environments (Slabbekoorn & Peet 2003; Mockford &
1
Present address: Department of Environment and Conservation, Great Southern Marshall 2009) and whales respond to low-frequency sonar
District, Katanning, WA, Australia.
2
emitted by the US navy by increasing the duration of their songs
Present address: Centre for Marine Futures, Oceans Institute, University of
Western Australia, Crawley, Perth, WA, Australia.
(Miller et al. 2000). Although relationships between the environ-
3
Deceased. ment and signal characteristics are best known for visual systems,

0003-3472/$38.00 Ó 2012 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.anbehav.2011.12.028
784 J. L. Kelley et al. / Animal Behaviour 83 (2012) 783e791

little attention has been paid to variation in luminance contrast, behaviour of fish in each of our treatment groups. Shoaling is the
which refers to differences in the amount of light reflected from an primary social behaviour for most teleost fishes and provides the
object compared with the background (Wyszecki & Stiles 1982). basis for key skills including foraging behaviour (e.g. food detec-
Animals such as reptiles, amphibians, cephalopods and fishes tion), reproduction (e.g. group spawning) and predator defence
have changeable coloration, allowing them to conceal their colora- (e.g. vigilance, attack dilution; Pitcher & Parrish 1993). Further-
tion for predator defence (Hanlon 2007) and/or increase their more, there is a body of experimental evidence linking both body
conspicuousness for communication with conspecifics (Stuart-Fox coloration (McRobert & Bradner 1998; Gomez-Laplaza 2009;
et al. 2008; Doucet & Mennill 2010). Such changes in coloration Rodgers et al. 2010, including our study species) and patterning
could also act to compensate for poor visual environments if, for (Engeszer et al. 2004; Peichel 2004; Rosenthal & Ryan 2005) with
example, the brightness of colour patches can be increased to shoaling behaviour. If changes in the composition and/or lumi-
enhance luminance contrast with the background. Examples of nance of light affect the expression (and conspicuousness) of colour
changes in skin coloration and body pattern for background patterns then we would expect that shoaling behaviour, which is
matching are well known, particularly for cephalopods (Allen et al. influenced by these traits, would also be affected. Specifically, we
2010) and fishes (Whiteley et al. 2009; Clarke & Schluter 2011), but predicted that fish might compensate for poor signalling environ-
the effects of environmental light (e.g. luminance and spectral ments (i.e. reduced luminance and/or availability of short wave-
composition) on the brightness of colour patterns and their use as length light) by increasing the expression of their colour patterns
behavioural communication signals is less clear. A notable exception and adjusting their shoaling behaviour.
is the diurnal colour change of a small freshwater fish (the neon tetra,
Paracheirodon innesi), in which the lateral stripe is an iridescent blue- METHODS
green at noon but fades to violet at night (Lythgoe & Shand 1983).
Freshwater habitats present particularly spectrally diverse Collection and Maintenance of Fish
environments because dissolved organic matter affects the scat-
tering, absorption and attenuation of light (Lythgoe 1979). While The western rainbowfish belongs to the family Melanotaenidae,
suspended particulate matter (e.g. mineral particles, plankton and a colourful group of freshwater fishes that is endemic to Australia
detritus) causes wavelength-independent scattering and attenua- and Papua New Guinea. The colour patterns of western rainbowfish
tion of light, dissolved organic matter tends to absorb short consist of one or two dark mid-lateral stripes, several narrower red/
wavelengths of light (ultraviolet-blue) selectively, resulting in orange lateral stripes and two or three black sublateral stripes. There
a shift in the wavelength of maximum transmission of the water is also a red/orange colour patch on the operculum (‘operculum
towards longer wavelengths (reds; Kirk 1977; Bowling et al. 1986). spot’) and the caudal fin can be colourless, yellow, orange or red.
Waters that are eutrophic or highly turbid also have similar spectral Previous research with this species has shown that colour patterns
properties to those environments rich in dissolved organic matter are highly variable among populations; variation in the area of
(Jerlov 1976; Bowling et al. 1986). pattern (size of the cheek spot and number of lateral stripes) is
While a number of studies have reported a relationship between related to the level of predation risk while colour pattern brightness
the properties of the water and fish coloration (Reimchen 1989; appears to be related to environmental parameters (e.g. water
Seehausen et al. 1997; Boughman 2001; Fuller 2002; Morrongiello velocity, turbidity, vegetation, temperature; Young et al. 2010).
et al. 2010), these have generally been field based making it diffi- The rainbowfish used in these experiments were collected from
 
cult to disentangle the effects of selection and phenotypic plasticity. Wittenoom Gorge (22 14.8350 S, 118 19.420 E), part of the Fortescue
Furthermore, increasing attention has been paid to the effects of River catchment which runs through the Hamersley Range in the
increased turbidity or eutrophication on mating behaviour and Pilbara region of Western Australia. Fish used to evaluate the effect
sexual selection (Candolin et al. 2007; Wong et al. 2007); under- of lighting on the area of colour patterns and shoaling behaviour
standing the plasticity of the colour signals that underlie these were collected in May 2006 for experiments in September 2007,
processes is important for predicting how aquatic communities while those used to examine changes in colour pattern reflectance
respond to anthropogenic impacts (Seehausen et al. 1997). were collected in January 2011 for use in May 2011. We expected
We adopted an experimental approach to test whether a col- any effects of captivity on fish coloration (e.g. from diet, lighting
ourful freshwater fish, the western rainbowfish, Melanotaenia conditions) to be the same for both cohorts. Other fishes occupying
australis, responds to changes in the light environment through this habitat include predators such as spangled perch, Leiopother-
alterations in the expression of its colour patterns and/or shoaling apon unicolor, and Fortescue grunters, Leiopotherapon aheneus, and
behaviour. The western rainbowfish is one of Western Australia’s a catfish, Hyrtl’s tandan, Neosilurus hyrtlii. Like other freshwater
most ubiquitous freshwater fish and occupies a huge range of light habitats in the Pilbara, for most of the year the river forms a series
environments including clear springs, creeks, pools and lakes with of pools that become connected during summer rainfall events
varying clarity and water colour (Allen et al. 2003). We simulated (Beesley & Prince 2010). During the dry period, changes in water
environments that are rich in organic matter using coloured filters quality result from the breakdown of organic matter and lack of
that reduce both the availability of short wavelength light and freshwater input or nutrient runoff increasing the algal content of
luminance (light intensity). Previous work with fish (black bream, the water (in agricultural areas). During the wet season, more
Acanthopagrus butcheri) has shown that these filters can induce dramatic changes in water quality can result as rainfall runoff leads
ontogenetic changes in the expression (i.e. sensitivity) of the visual to high sedimentation and reduced water clarity.
pigments (Shand et al. 2008). Neutral-density filters were used in Fish were maintained in aquaria (66  36 cm and 45 cm high) in
a second treatment group to reduce luminance without affecting mixed-sex groups (approximately 1:1 sex ratio) and fed a diet of
spectral composition. A third control group was not exposed to any commercial flake food at the University of Western Australia.
change in the light environment. Following 2 weeks in the treat- Aquarium lighting was provided by overhead fluorescent 30 W
ment tanks, the area and brightness of fish colour patterns were bulbs on a 12:12 h light:dark cycle and the water temperature was
measured using digital photography and spectrometry, respec- maintained at 26  1  C. As all experiments were performed
tively, and compared among light treatments. We also examined indoors no ultraviolet light was present. Rainbowfish are not
the combined effects of changes in the light environment and fish strongly sexually dimorphic but males often have brighter patterns
colour pattern expression by observing the social (shoaling) than females (Allen et al. 2003). In fishes, females often show
J. L. Kelley et al. / Animal Behaviour 83 (2012) 783e791 785

a stronger shoaling tendency than males (because of differences in habitat of the western rainbowfish are shown in Fig. 1b for
reproductive potential between the sexes, Magurran et al. 1992); comparison. These measures were made at the collection site of the
thus we decided to focus on females only for our experiments. fish (Fortescue River upstream, a relatively clear water site) and
a downstream site of the same river (containing greater amounts of
 
Simulating Differences in the Light Environment dissolved organic matter; 21 17.5480 S, 116 08.6720 E) in May 2010. In
the experiment, tank treatments were randomly distributed
We used acetate filters to simulate changes in the light envi- spatially to account for lighting variations among aquaria; however,
ronment because these affect the wavelength composition and we also measured the total absolute irradiance of light in each tank
luminance of light entering the water without altering the prop- so that this could be included as part of our analysis. Brown card-
erties of the water (e.g. light scattering from suspended particles). board was placed on the sides of all tanks to prevent fish in adjacent
Environments rich in organic matter (‘yellow filter’ treatment treatment tanks from interacting with one another. Each treatment
group) were simulated by placing two layers of yellow acetate filter tank contained a layer of brown gravel, an airstone and a piece of
(‘764 Sun Colour Straw’, Lee Filters) around the front, back and top artificial plant. Sample sizes used in each part of the experiment are
of each tank. Like natural dissolved organic matter, these filters given in the relevant sections below.
acted to reduce the intensity and availability of short wavelength
light (transmittance between 400 and 550 nm ¼ 30e80%, see Effect of Treatment on Colour Pattern Area
Shand et al. 2008). The ‘neutral-filter’ treatments were set up in the
same way but using two layers of neutral-density filter (0.3ND Prior to being assigned to their treatments, each female was
filter, Lee Filters; approximately 50% transmission 400e700 nm); lightly anaesthetized in clove oil (80 mg/litre) and photographed
these acted to reduce luminance without affecting wavelength against a white standard using a digital camera (Olympus Mju
composition. This allowed us to disentangle the effects of changes 1030SW). Lighting was provided by two bench lamps containing
in light wavelength (colour) and luminance (light brightness) on broad-spectrum fluorescent tubes (2  11 W each). It was not
fish colour patterns and social behaviour. A final treatment group possible to set the white balance on our camera manually so we
(‘control’) was set up in the same way without any filters and using used the factory setting for photography under fluorescence
a transparent glass lid. lighting. We also included a code for each individual’s photograph
The absolute irradiance (total downwelling light measured over such that later analysis (see below) could be performed blind of

180 ) was measured for each treatment group using a spectrometer treatment group. Anaesthesia is known to affect the brightness of
(see below and Fig. 1a). Measures of absolute irradiance in a natural fish colour patterns (Gray et al. 2011) and we have noticed that
clove oil causes a progressive darkening of the skin, particularly the
dorsal surface. The Animal Ethics Committee (AEC) of Western
3.5E.07 Australia currently advises the use of clove oil for fish anaesthesia;
(a)
nevertheless, colour patterns should be affected equally across all
3.0E.07 treatment groups as each fish spent the same period of time in
2.5E.07 clove oil (<1 min). Each fish was allowed to recover in conditioned
Absolute irradiance (µmoles photons/m2 per s per nm)

(Armour Coat) aerated water before being assigned to a treatment


2.0E.07 group. To avoid stress associated with testing fish in isolation,
females were placed in their treatment tanks in groups of four,
1.5E.07 giving a total of 72 fish allocated to this part of the experiment
(N ¼ 24 fish/treatment). However, data from two females were
1.0E.07
missing because of photography error, giving a final sample size of
0.5E.07 70 individuals (yellow filter: N ¼ 23; neutral filter: N ¼ 24; control:
N ¼ 23). Fish were maintained in their treatment groups for just
0 over 2 weeks (15 days), after which they were rephotographed
400 500 600 700
before taking part in the shoaling trials.
0.008 (b) Each fish photograph was imported into Image J, version 1.44
(http://rsb.info.nih.gov/ij/index.html) and calibrated for size
0.007 according to the ruler photographed with each image. Images were
0.006 first split into separate RGB (red, green, blue) colour channels, each
displayed as an 8-bit greyscale image, to reveal different compo-
0.005
nents of the colour pattern (Stevens et al. 2007). A preliminary
0.004 inspection revealed that the green channel appeared to show most
detail of the colour patterns, while the blue channel accentuated
0.003
the orange coloration such as the lateral stripe and operculum spot.
0.002 We therefore used the green and blue channels for different
0.001 elements of our subsequent analyses. We selected a threshold value
of 60 for the green channel and for each fish we calculated the total
0 area of pigmentation on the body (excluding fins) that lay between
300 400 500 600 700
the values of 0 and 60. In the blue channel, we selected a threshold
Wavelength (nm)
of 10 before measuring the area of the operculum spot and the area
Figure 1. (a) Absolute irradiance (mmoles photons/m2 per s per nm) measured in the comprising the orange lateral stripes (i.e. total area of these
three treatment groups: orange filter (dotted line), neutral filter (solid black line) patterns with pixel intensities between 0 and 10). Thus we used the
and control (grey line; no filter) treatment tanks. (b) Absolute irradiance (mmoles photographs to measure the shape and size of the colour patterns,
photons/m2 per s per nm) measured in the upper (dashed line) and lower (solid line)
reaches of the Fortescue River. Note that the spectral composition of light at the
while spectrometry (see below) was used to evaluate the colour
downstream site (containing more dissolved organic matter) is shifted towards longer and brightness of these patches. We chose these regions of color-
wavelengths compared with the upstream (clear spring) site. ation based on previous studies showing population variation in
786 J. L. Kelley et al. / Animal Behaviour 83 (2012) 783e791


these traits (Young et al. 2010). We also measured the standard another, with each fibre at 45 to the fish’s body. One of these fibres,
body length (SL) of each fish to determine the relationship between the ‘read’ fibre (400 nm), was fitted with a collimating lens focused
colour patterns and body length. to sample an area of 1 mm diameter. The read fibre was connected
to a USB4000 spectrometer (Ocean Optics, Inc., Dunedin, FL, U.S.A.),
Effect of Treatment on Shoaling Behaviour configured to sample light wavelengths within the range of
172e881 nm. The other optical fibre (600 nm) was connected to
Shoaling trials were performed in a circular observation area a light source (200e1700 nm; Mini-DT-1000, Analytical Instru-
(diameter ¼ 70 cm, depth ¼ 7 cm), which had black sides, and ment Systems, Inc., Flemington, NJ, U.S.A.). Prior to each fish being
a white base marked with 5 cm grid lines. Lighting was provided by measured, we ensured that the probes were in alignment and took
two fluorescent bench lamps (2  11 W bulbs each) positioned on measurements of white (WS-1, Ocean Optics, Inc.: 100% reflec-
each side of the arena. The shallow water allowed us to observe tance) and dark standards, ensuring that the height of the stage was
shoaling behaviour in two dimensions, which is advantageous for optimal for maximum reflectance. The reflectance measures for
evaluating interindividual shoaling distance. We covered the top of each fish were therefore performed in relation to these light/dark
the shoaling arena with a filter, such that fish were tested under the standards, with the software (Spectrasuite, Ocean Optics, Inc.) set to
same lighting treatments that they had experienced during the average across three sample spectra.
previous 15-day period. Thus yellow-filter fish were tested with Fish were oriented with their right side down such that the
two layers of yellow filter, neutral-filter fish were tested with two illuminating fibre was positioned to provide light from the anterior
layers of neutral filter and control fish were tested with transparent direction of the fish while the read fibre was positioned to collect
acetate. Glass was too heavy for the apparatus and differences in light from the posterior end of the fish. Ideally, illumination would
the transmittance of glass and acetate occur in the ultraviolet be provided from the direction of ambient light in nature (i.e.
wavelengths, which were absent during the experiment. dorsally, Endler 1990) but this proved difficult with the curvature of
Pairs of females originating from the same treatment group the fish’s body. We sampled along the predominant mid-lateral
were placed in a circular shoaling arena and left to acclimate for orange stripe (posterioreanterior direction, along 1/3 of body),
10 min. We used pairs of females rather than groups of four to sampling approximately every one or two scales to give 5e10
increase the number of replicates that were performed. Although spectral samples per individual. The fish was moved up and down
this could lead to the development of social relationships (e.g. in the vertical plane to find the peak reflectance for each reading.
dominance interactions) that might not be observed in larger Following spectrometry, fish were allowed to recover in condi-
natural shoals, this was unlikely to occur during the short time tioned, aerated water before being returned to stock aquaria where
frame in which the fish were associated (15 min) but could have they played no further part in the experiment. For each sampling
occurred during the 15-day treatment. Furthermore, simple pair- point, 3648 spectral readings were obtained across the range
wise interactions between individuals can provide reliable indica- 172e881 nm. These data were reduced to the interval 300e700 nm
tors of the spatial organization of larger (e.g. 10 and 30 fish) social using moving averages across 2 nm intervals (every 10 readings) to
groups (Katz et al. 2011). We used all 72 females that had been give 259 data points for each reflectance measure (i.e. a measure of
allocated to the light treatments, to give a sample size of 36 pairs of reflectance for each of the 259 different wavelengths). We then
fish (N ¼ 12 pairs per treatment). Following the acclimation period, calculated a mean reflectance spectrum for each individual fish
we recorded the shoaling behaviour of the fish for a period of 5 min and subsequent statistical analyses (see below) are based on each
using a digital video camera (Canon MVX150). Following each trial, individual’s mean reflectance spectrum.
fish were returned to stock aquaria where they played no further For the irradiance data, a 600 nm optical fibre was fitted with
part in these experiments. Videos were imported into the program a cosine collector (CC-3-UV, Ocean Optics, Inc.) and oriented
Zoom Player and the distance between shoaling fish analysed using vertically (to collect downwelling light) in each treatment tank.
the program Pesci (M. Martinelli, University of Padova, Italy). Absolute irradiance measures (in mmoles of photons/nm) were
Distance was calibrated using the markings on the bottom of the performed with reference to a calibration lamp (LS-1-CAL-220,
arena and readings were taken at 30 s intervals to give a total of 10 300e1050 nm; Ocean Optics, Inc.) and measured in Spectrasuite.
readings, which were averaged for each pair of fish sampled. Total absolute irradiance (in photons/cm2 per s) was calculated for
each treatment tank by integrating irradiance over the wavelength
Effect of Treatment on Colour Pattern Brightness range 400e700 nm (no ultraviolet light, 300e400 nm, was present
in our laboratory experiment).
Following the methods described above, a separate group of
female fish was randomly assigned to yellow-filter, neutral-filter or Statistical Analysis
control treatment groups, where they were maintained for a period
of 15 days in groups of four. A total of 64 females were used, none of We used general linear mixed models (GLMMs) to test the effect
which took part in the previous experiment. Twelve individuals of treatment (yellow filter, neutral-density filter or control) on our
were removed and placed in stock aquaria because they were three response variables: the total area of pigmentation, the area
behaving aggressively or were subjected to aggression from other comprising the lateral orange stripes and the area of pigmentation
fish (see Ethical Note, below); this gave us a sample size of 52 in the operculum patch (all measured following the 15 days in
females for the spectrometry (yellow filter: N ¼ 17; neutral filter: treatment tanks). We first confirmed that there were no differences
N ¼ 16; control: N ¼ 19). Fish were removed from their treatment in patterns between groups before they were assigned to their
tanks in their groups of four, assigned an identification code (such treatments. Individual tanks (N ¼ 6 per treatment) were allocated
that subsequent data analysis was performed blind of treatment) to each treatment; we therefore entered tank nested within
and transferred to a dark room for the spectrometry. treatment as a random factor. Thus the degrees of freedom reflect
Fish were anaesthetized using the methods described above the number of tanks rather than the number of individual fish. We
before being placed on a matt black wedge, allowing the body to be also entered ‘standard body length’ as a covariate to determine its
positioned horizontally. The wedge was placed on a photography effect on the dependent variables. Although our response variables
stage that allowed fine movement in the vertical plane. Laboratory were weakly correlated with one another (correlation coefficients

clamps were used to position two optical fibres at 90 to one ranging from 0.22 to 0.37), we chose to analyse them separately
J. L. Kelley et al. / Animal Behaviour 83 (2012) 783e791 787

because a preliminary principal components analysis (PCA) showed Table 1


that they were equally loaded in different directions, making the Results of general linear mixed models, testing the effect of light treatment (three
levels) and standard body length (SL) on the three pattern (response) variables
results difficult to interpret. All statistical analyses were performed
in JMP version 9.0 (SAS Institute Inc., Cary, NC, U.S.A.). Response variable Fixed effect df F P
GLMMs were also used to test for the effect of treatment on colour Total area of pigmentation Treatment 2, 13.8 1.93 0.182
pattern brightness. We first used PCA to reduce the 259 wavelength SL 1, 64.1 17.01 <0.0001
Area of striped pigmentation Treatment 2, 15.1 4.02 0.040
measures to two principal components; these typically describe the
SL 1, 65.8 5.82 0.019
brightness (PC1 if loaded equally across all wavelengths) and hue Area of operculum pigmentation Treatment 2, 13.3 0.94 0.413
(PC2) of colour patterns, respectively (Cuthill et al. 1999). In the SL 1, 62.9 0.11 0.737
absence of any information on the visual sensitivity of the study Tank was nested in treatment (six tanks per treatment) and entered as a random
species, PCA is a better alternative than the ‘segment classification’ factor. Significant effects are shown in bold.
method, which is unsuitable for species with more than three
photoreceptor types (Endler 1990); preliminary work with the
western rainbowfish has revealed five photoreceptor classes so far RESULTS
J.L. Kelley, N.S. Hart, J. Shand, S.P. Collin & D.M. Hunt, unpublished
data). We accounted for differences in ambient lighting (caused by Effect of Treatment on Colour Pattern Area
tank position) in our analysis by entering the total absolute irradiance
for each tank as a random factor. This was particularly important We confirmed that there was no difference between individuals
because, despite our attempts to equalize luminance across the in the total area of pigmentation (F2, 68 ¼ 0.065, P ¼ 0.937), the area
treatments, control tanks had higher total absolute irradiance than of stripe pigmentation (F2, 68 ¼ 0.50, P ¼ 0.611) or the area of
neutral tanks (control mean ¼ 6.16e13 photons/cm2 per s, range operculum pigmentation (F2, 68 ¼ 0.15, P ¼ 0.866) before fish were
3.11e13 photons/cm2 per s, neutral mean ¼ 4.49e13 photons/cm2 per assigned to their treatment groups. Treatment had no effect on the
s, range 1.1e13 photons/cm2 per s; ANOVA: F2, 13 ¼ 4.95, P ¼ 0.025; total area of pigmentation but a significant effect on the area of
post hoc Tukey test: P < 0.05). However, there was no significant striped pigmentation (Table 1, Fig. 2); a post hoc Tukey test revealed
difference between the total absolute irradiance in the neutral-filter that fish in the yellow-filter group displayed a greater area of stripe
and yellow-filter treatments (P < 0.05). A total of 52 reflectance pigmentation than those in the control group (Fig. 2; P < 0.05).
spectra (based on a mean value for each individual) were used in the There was no effect of treatment on the area of operculum
final analysis. Fish body length (SL) data were only collected for pigmentation (P > 0.05; Table 1). Both the total area of pigmenta-
a subset of 18 fish; these data were therefore used to investigate the tion (t68 ¼ 4.15, P < 0.001) and area of striped pigmentation
relationship between colour (hue and brightness) and body size. (t68 ¼ 2.22, P ¼ 0.030) showed a positive linear relationship with
We tested for the effects of treatment on shoaling behaviour fish standard body length (Table 1).
using ANCOVA to control for variation in body length, which can
influence shoaling behaviour in fishes (Ranta et al. 1993). As we did Effect of Treatment on Colour Pattern Brightness
not determine which individuals were selected for the shoaling
trials we used the variance in body length (variance in standard PC1 showed strong positive loadings across all wavelengths
length for the four individual fish in each treatment tank) as a co- (300e700 nm; loadings: 0.80e1.0), whereas PC2 was weakly
variate in the analysis. positively loaded across the range 350e500 nm (loadings: 0.3e0.5)
and weakly negatively loaded from 550 to 700 nm (loadings: 0.1
Ethical Note to 0.23). The two principal components that were extracted (PC1
and PC2) explained 84.1% and 10.5% of the variation in our data,
This work was performed in accordance with the Australian respectively. For PC1, which is usually correlated with mean
Code of Practice for the Care and Use of Animals for Scientific reflectance (or brightness) there was a significant effect of treat-
Purposes (2004) and approved by the Animal Ethics Committee of ment group (F2, 19.5 ¼ 3.98, P ¼ 0.036); a post hoc Tukey test
the University of Western Australia (licence RA/3/100/691). In the revealed a significant difference between the control and yellow-
field, fish were captured using a 3 m seine net and adults placed in filter groups at P < 0.05 (Figs 3, 4a).
sealed plastic bags for transport by air to the University. The
transport bags contained river water, a small amount of vegetation,
Armour Coat (for skin protection) and zeolite minerals (for
10
ammonia absorption) and were packed into polystyrene crates.
During the 15-day laboratory experiment (light treatments),
8
Stripe area (mm2)

artificial plants were used to provide refuge for the fish. Fish were
monitored twice daily for signs of aggression, which included
6
nipping and chasing behaviours. Fish that were aggressive towards
others or were subjected to aggression (i.e. showed signs of fin
4
damage) were removed and isolated (for recovery) before being
returned to stock aquaria. A total of 12 fish were excluded from our 2
experiments because of aggressive behaviours and we did not
observe any other adverse behaviour in our treatment tanks. 0
During the photography and spectrometry we minimized the time Y N C
fish spent in anaesthesia (which was <1 min and approximately Treatment group
3 min, respectively) and ensured rapid recovery by placing fish in
aerated water after the procedure. Fish did not show any signs of ill Figure 2. Differences in the mean 1 SE area of pigment (mm2) comprising the two
lateral orange stripes (revealed by thresholding digital images, see Methods) in rain-
health during the photography or spectrometry and no deaths bowfish assigned to the yellow-filter (Y), neutral-filter (N) and control (no filter, C)
occurred as a result of these procedures. At the end of the experi- treatment groups. Sample sizes for the Y, N and C treatment groups were 23, 24 and 23
ments, all fish were returned to stock aquaria. females, respectively.
788 J. L. Kelley et al. / Animal Behaviour 83 (2012) 783e791

There was also an effect of treatment on PC2 (typically correlated 20


(a)
with hue; F2, 19.5 ¼ 4.06, P ¼ 0.034; Fig. 4b); a post hoc Tukey test
revealed a significant different between the yellow- and neutral-

Least – squares mean (pc1)


15
filter groups (P < 0.05). Thus fish in the yellow-filter treatment
with positive PC2 scores tended to have colours with high reflec-
10
tance at short wavelengths (violet-blue: 400e500 nm) and low
reflectance at long wavelengths (yellow-orange: >550 nm) while
5
the reverse was true for fish in the neutral-filter treatment (high
reflectance at long wavelengths, low reflectance at short wave-
lengths). However, PC2 accounted for only 10.5% of the variation in 0
our data and the average spectrum for each treatment group (Fig. 3)
suggests changes in brightness rather than hue. The random effect of –5
total absolute irradiance had a significant effect in our models (PC1:
F13, 36 ¼ 3.54, P ¼ 0.001; PC2: F13, 36 ¼ 7.60, P < 0.001). There was no –10
relationship between body length (SL) and PC1 (t16 ¼ 0.24, P ¼ 0.81) Y N C
or PC2 (t16 ¼ 0.94, P ¼ 0.36).
6
(b)
Effect of Treatment on Shoaling Behaviour

Least – squares mean (pc2)


4
We found a significant effect of treatment on shoaling behaviour
(F2, 32 ¼ 3.43, P ¼ 0.045; Fig. 5) but no effect of the covariate (tank 2
variance in body length: F1, 32 ¼ 0.48, P ¼ 0.49). Following removal
of the covariate from our model (treatment: F2, 33 ¼ 3.87, 0
P ¼ 0.031), a post hoc Tukey test revealed a significant difference
between the yellow-filter treatment and both the neutral-filter and –2
control treatment groups (P < 0.05; Fig. 5). Fish in the yellow
treatment group retained an average distance of 22.8 þ 3.1 mm –4
(mean þ SE) from their shoaling partner while those in the neutral
and control groups remained closer together (15.0 þ 1.5 mm and
–6
14.5 þ 2.2 mm, respectively). Y N C

DISCUSSION Figure 4. Least-squares means  1 SE from general linear mixed models used to test
the effect of treatment (yellow filter, Y, N ¼ 17; neutral filter, N, N ¼ 16; control group,
C, N ¼ 19) on the response variables (a) PC1 and (b) PC2. PC1 represents brightness
While links between the light environment and a species’ colour (achromatic reflectance) while PC2 represents variation in the relative reflectance of
patterns are often considered in the context of signal evolution, we the lateral stripe at short (350e500 nm) versus long (550e700 nm) wavelengths;
have provided evidence that changes in the brightness of colour positive PC2 scores suggest colour pigments that reflect strongly at short wavelengths
(e.g. blues) while negative PC2 scores represent patches that are more reflective at
patches can also arise through alterations in the light environment.
longer wavelengths (e.g. oranges).
We found that a reduction in the relative availability of short
wavelength light caused an increase in the brightness (reflectance)
of colour patterns. Fish responded to the combined effects of relative availability of short wavelength light had greater interin-
changes in their coloration and the ambient light environment by dividual shoaling distances (shoaled further apart) than those in
altering their shoaling behaviour; fish in waters with reduced the low luminance (neutral filter) or control groups. Our findings
suggest that alterations in the light environment affect the area and
brightness of colour patterns, which could subsequently influence
45
their use as behavioural signals.
40
35
Mean % reflectance

30 30
25
Mean distance (mm)

25
20
20
15
15
10
5 10

0 5
300 400 500 600 700
Wavelength (nm) 0
Y N C
Figure 3. Mean percentage reflectance of the upper orange lateral stripe of fish Treatment group
assigned to the yellow-filter (solid line), neutral-filter (dashed line) and control (dotted
line) treatment groups. Standard errors surrounding the mean are shown in grey (1 Figure 5. Effect of yellow (Y), neutral (N) and control (C) treatment groups (N ¼ 12
SE). Sample sizes are 17, 16 and 21 fish for the yellow-filter, neutral-filter and control pairs per treatment) on the mean interindividual distance between shoaling females
groups, respectively. (error bars represent 1 SE).
J. L. Kelley et al. / Animal Behaviour 83 (2012) 783e791 789

Previous studies have investigated how differences in the light conditions have demonstrated changes in the optomotor response
environment contribute to variation in the hue (colour) and (Kroger et al. 2003) and in the expression of opsin genes (which
brightness of animal colour pattern signals. For example, bird determine the sensitivity of the visual pigments; Shand et al. 2008)
species from closed habitats that are characterized by a predomi- in response to changes in the light environment.
nance of long wavelength light tend to exhibit colours that are rich In this study, the red lateral stripes that were measured are
in long wavelengths (McNaught & Owens 2002), which should located in the mid- to mid-ventral posterior regions of the body and
serve to enhance chromatic contrast against background light are therefore less likely to be affected by upwelling light than
(Endler 1990). Colour patch brightness has also been shown to vary ventral colour patterns. Although we understand little of the
across species; birds in open canopy habitats rich in blue and UV function of colour patterns in rainbowfishes, studies with other
light tend to exhibit brighter coloration (higher reflectance relative species have revealed that lateral longitudinal stripes can serve as
to incident light) than those occupying understory environments important cues in shoaling behaviour (Peichel 2004; Rosenthal &
that are rich in green and orange light; these differences in color- Ryan 2005). These stripes must appear conspicuous against the
ation may allow birds to appear cryptic in their respective natural background of the fish’s skin if they are to serve a communicative
environments (Gomez & Thery 2004). In dwarf chameleons, Bra- function. For example, Long & Houde (1989) examined female
dypodion spp., the pattern is similar with males in closed habitats preference for males when viewed behind orange filters (peak
exhibiting less bright coloration than those in open habitats, but transmittance 600e690 nm range) and found that the female’s
their colour patches allow for greater brightness contrast against usual preference for orange diminished under orange light,
the background in these closed habitats (Stuart-Fox et al. 2007). because of a lack of contrast of orange spots with the
These studies illustrate the importance of considering both surrounding skin.
ambient light and the (chromatic/luminance) contrast provided by Several studies have examined the relationship between
the visual background (Endler 1990; Stuart-Fox et al. 2007). colour patterns and the visual sensitivity of the receiver; for
Field studies with other species of fish have also reported example in guppies, Poecilia reticulata, there is no relationship
a correlation between the light environment and red body colora- between male coloration and water colour, but females from
tion. For example, populations of southern pygmy perch, Nanno- waters with greater relative long wavelength transmittance show
perca australis, tend to have redder coloration in habitats that are a stronger preference for orange and black males (Endler & Houde
rich in long wavelengths (550e700 nm; Morrongiello et al. 2010) 1995). In studies with sticklebacks, males inhabiting waters
and in red shiners, Cyprinella lutrensis, males from turbid habitats characterized by the dominance of long wavelength light tend to
have brighter caudal fins during the breeding season than those exhibit less red coloration which is linked to the females’ reduced
from less turbid waters (Dugas & Franssen 2011). In contrast, sensitivity to red in these populations (Boughman 2001). The light
sticklebacks, Gasterosteus aculeatus, from stained waters that have environment has also been shown to influence mating behaviour
low transmittance in the 350d450 nm range have less red throat in guppies with females responding more to males observed
pigmentation than those from clearer waters or tend to be replaced under simulated woodland shade environments (which stimu-
by melanic (black) males that may be more conspicuous in these lates cones sensitive to short and medium wavelengths) than
environments (Reimchen 1989). The ventral red throat pouches in under forest shade or early/late environments (which stimulate
sticklebacks from stained habitats are unlikely to be perceived as cones sensitive to long and mediumelong wavelengths, respec-
conspicuous because the long wavelength-shifted upwelling light tively; Gamble et al. 2003).
and the high reflectance of ventral surfaces minimize contrast with Our observation that shoaling behaviour was only modified in
the background (Reimchen 1989). the yellow-filter treatment group (reduced relative availability of
In mid-water-dwelling fishes, such as rainbowfishes, individuals short wavelength light) could suggest that the increase in area and/
will often be viewed against the background provided by the water or brightness of colour patterns was sufficient for these fish to
column, primarily consisting of sidewelling light. Colour patches interact over greater distances, but not sufficient to alter the
will be perceived as brighter if they reflect wavelengths that are behaviour of fish in the other (neutral and control) groups. The
relatively more abundant in the downwelling light than the colour pattern area and brightness of fish in the neutral group
sidewelling light (Endler 1990); thus red colour patterns should be appeared to be intermediate to those in the control and yellow-
most conspicuous in habitats that are rich in dissolved organic filter groups, providing some evidence for this. An alternative
matter (Lythgoe 1979; Levine et al. 1980). The enhanced brightness explanation for the differences in shoaling behaviour between
of red colour patterns observed in our reduced light intensity and treatment groups is that changes in the wavelength composition of
red-shifted (yellow filter) light environments should act to increase the light stimulated some other change in the behaviour of female
the brightness of the red-coloured stripes, providing enhanced rainbowfish. Changes in shoaling density (i.e. interindividual
(luminance) contrast against a reduced light intensity background. shoaling distances) are often associated with the trade-off between
However, the visual background provided by the surrounding skin foraging efficiency and predation risk with an increase in shoal
pigmentation (which is silvery and highly reflective in the ultravi- density facilitating increased protection from predators (e.g. risk
olet (300e400 nm) wavelength range) is also important and, in this dilution) at the cost of reduced foraging efficiency (because of
case, the increased brightness of the red stripes in red-shifted light competition among group members; Pitcher & Parrish 1993). Shifts
environments would produce less luminance contrast against the in the properties of light that are consistent with diurnal variations
silvery skin (assuming the reflectance of these patterns remained may induce behavioural shifts, or light conditions that are consis-
unchanged), reducing the overall conspicuousness of the entire tent with high turbidity may be associated with a reduction in
colour patterns. Information on the visual sensitivity of rainbowfish predation risk and an increased motivation to feed (reviewed in
would allow us to reveal whether the observed colour pattern Utne-Palm 2002).
changes facilitate enhanced conspicuousness or crypsis of the We acknowledge that linking shoaling behaviour and fish
colour pattern components (see concluding paragraph also). We coloration to environmental light, rather than addressing a more
acknowledge that the spectral composition of the fluorescent specific behaviour, such as female choice for male colour patterns,
lighting used in this study is very different to the spectral charac- makes it difficult to predict how these variables might interact
teristics of light in the species’ natural habitat (Fig. 1). None the less, to influence animal fitness. However, many studies have linked
other studies conducted under similar (artificial) lighting red/orange coloration to fitness traits in freshwater fishes; for
790 J. L. Kelley et al. / Animal Behaviour 83 (2012) 783e791

example, by selecting males with high levels of red/orange color- Cuthill, I. C., Bennett, A. T. D., Partridge, J. C. & Maier, E. J. 1999. Plumage
reflectance and the objective assessment of avian sexual dichromatism. Amer-
ation female sticklebacks avoid parasitized males (and may
ican Naturalist, 153, 183e200.
produce offspring with increased capacity for parasite resistance; Doucet, S. M. & Mennill, D. J. 2010. Dynamic sexual dichromatism in an explosively
Milinski & Bakker 1990) and female guppies produce offspring with breeding Neotropical toad. Biology Letters, 6, 63e66.
enhanced antipredator skills (i.e. ‘good genes’ models of sexual Dugas, M. B. & Franssen, N. R. 2011. Nuptial coloration of red shiners (Cyprinella
lutrensis) is more intense in turbid habitats. Naturwissenschaften, 98,
selection; Evans et al. 2004). We hope our work will stimulate other 247e251.
researchers to consider how environmentally mediated colour Endler, J. A. 1987. Predation, light intensity and courtship behavior in Poecilia
pattern plasticity may influence evolutionary processes such as reticulata (Pisces: Poeciliidae). Animal Behaviour, 35, 1376e1385.
Endler, J. A. 1990. On the measurement and classification of colour in studies of
sexual selection. animal colour patterns. Biological Journal of the Linnean Society, 41, 315e352.
We cannot predict whether the increased reflectance (bright- Endler, J. A. 1991. Variation in the appearance of guppy color patterns to guppies
ness) of rainbowfish colour patterns serves to increase conspicu- and their predators under different visual conditions. Vision Research, 31,
587e608.
ousness against the background (light) environment without Endler, J. A. 1992. Signals, signal conditions and the direction of evolution. American
considering the visual sensitivity of the rainbowfish (or its preda- Naturalist, 139, S125eS153.
tors). While we are in the process of describing the visual system of Endler, J. A. & Basolo, A. L. 1998. Sensory ecology, receiver biases and sexual
selection. Trends in Ecology & Evolution, 13, 415e420.
the western rainbowfish, a member of the same genus, Melano- Endler, J. A. & Houde, A. E. 1995. Geographic variation in female preferences for
taenia maccullochi, has six spectral classes of cone with maximum male traits in Poecilia reticulata. Evolution, 49, 456e468.
sensitivities ranging from the ultraviolet (lmax ¼ 377 nm) to Engeszer, R. E., Ryan, M. J. & Parichy, D. M. 2004. Learned social preference in
yellows/oranges (lmax ¼ 577 nm; Reckel et al. 2002), which zebrafish. Current Biology, 14, 881e884.
Evans, J. P., Kelley, J. L., Bisazza, A., Finazzo, E. & Pilastro, A. 2004. Sire attrac-
provides a wide spectral range for visually mediated behaviours. tiveness influences offspring performance in guppies. Proceedings of the Royal
With data on the sensitivity of the photoreceptors in each cone Society B, 271, 2035e2042.
class for M. australis, it would be interesting to use visual models to Fuller, R. C. 2002. Lighting environment predicts the relative abundance of male
colour morphs in bluefin killifish (Lucania goodei) populations. Proceedings of
determine whether the increased colour pattern reflectance acts to the Royal Society B, 269, 1457e1465.
maintain communication in a poor signalling environment. These Gamble, S., Lindholm, A. K., Endler, J. A. & Brooks, R. 2003. Environmental vari-
types of models could be used to determine the distances over ation and the maintenance of polymorphism: the effect of ambient light
spectrum on mating behaviour and sexual selection in guppies. Ecology Letters,
which individuals can maintain shoaling associations in variable 6, 463e472.
light environments. Such an approach would allow deeper insights Gomez, D. & Thery, M. 2004. Influence of ambient light on the evolution of colour
into how changes in the aquatic light environment affect the signals: comparative analysis of a Neotropical rainforest bird community.
Ecology Letters, 7, 279e284.
physiology, social behaviour and ecology of fishes. Gomez-Laplaza, L. M. 2009. Recent social environment affects colour-assortative
shoaling in juvenile angelfish (Pterophyllum scalare). Behavioural Processes, 82,
39e44.
Acknowledgments Gray, S. M., Hart, F. L., Tremblay, M. E. M., Lisney, T. J. & Hawryshyn, C. W. 2011.
The effects of handling time, ambient light, and anaesthetic method, on the
We thank Suzanna Chan for her help with data collection and standardized measurement of fish colouration. Canadian Journal of Fisheries and
Aquatic Sciences, 68, 330e342.
Rick Roberts, Husnan Ziadi and Cameron Duggin for practical Hanlon, R. 2007. Cephalopod dynamic camouflage. Current Biology, 17, R400eR404.
assistance. Thanks also to Jon Evans for use of his spectrometer, Heindl, M. & Winkler, H. 2003. Vertical lek placement of forest-dwelling manakin
Dale Roberts for access to laboratory facilities and Bob Black for species (Aves, Pipridae) is associated with vertical gradients of ambient light.
Biological Journal of the Linnean Society, 80, 647e658.
statistical advice. Justin Marshall and Martin Stevens provided Jerlov, N. G. 1976. Marine Optics. Amsterdam: Elsevier.
advice on measuring fish coloration. We are especially grateful to Katz, Y., Tunstrom, K., Ioannou, C. C., Huepe, C. & Couzin, I. D. 2011. Inferring the
John Endler for numerous discussions about rainbowfish and for structure and dynamics of interactions in schooling fish. Proceedings of the
National Academy of Sciences, U.S.A., doi:10.1073/pnas. 1107583108, Published
suggesting the spectrometry part of this study. Nathan Hart, John online 27 July 2011.
Endler and the anonymous referees kindly provided excellent Kirk, J. T. O. 1977. Attenuation of light in natural waters. Australian Journal of Marine
comments that greatly improved the manuscript. J.L.K. is grateful to and Freshwater Research, 28, 497e508.
Kroger, R. H. H., Knoblauch, B. & Wagner, H.-J. 2003. Rearing in different photic
the University of Western Australia for a Postdoctoral Research
and spectral environments changes the optomotor response to chromatic
Fellowship. stimuli in the cichlid fish Aequidens pulcher. Journal of Experimental Biology, 206,
1643e1648.
Leal, M. & Fleishman, L. J. 2004. Differences in visual signal design and detect-
References ability between allopatric populations of Anolis lizards. American Naturalist, 163,
26e39.
Allen, G. R., Midgley, S. H. & Allen, M. 2003. Field Guide to the Freshwater Fishes of Levine, J. S., Lobel, P. S. & MacNicol, E. F. 1980. Visual communication in fishes. In:
Australia. Perth: Western Australian Museum. Environmental Physiology of Fishes (Ed. by M. A. Ali), pp. 447e475. New York:
Allen, J. J., Mathger, L. M., Barbosa, A., Buresch, K. C., Sogin, E., Schwartz, J., Plenum.
Chubb, C. & Hanlon, R. T. 2010. Cuttlefish dynamic camouflage: responses to Long, K. D. & Houde, A. E. 1989. Orange spots as a visual cue for female mate choice
substrate choice and integration of multiple visual cues. Proceedings of the Royal in the guppy (Poecilia reticulata). Ethology, 82, 316e324.
Society B, 277, 1031e1039. Lythgoe, J. N. 1979. The Ecology of Vision. Oxford: Clarendon Press.
Beesley, L. S. & Prince, J. 2010. Fish community structure in an intermittent river: Lythgoe, J. N. & Shand, J. 1983. Diel colour changes in the neon tetra Paracheirodon
the importance of environmental stability, landscape factors and within-pool innesi. Environmental Biology of Fishes, 8, 249e254.
habitat descriptors. Marine and Freshwater Research, 61, 605e614. McNaught, M. K. & Owens, I. P. F. 2002. Interspecific variation in plumage colour
Boughman, J. W. 2001. Divergent sexual selection enhances reproductive isolation among birds: species recognition or light environment? Journal of Evolutionary
in sticklebacks. Nature, 411, 944e948. Biology, 15, 505e514.
Boughman, J. W. 2002. How sensory drive can promote speciation. Trends in McRobert, S. P. & Bradner, J. 1998. The influence of body coloration on shoaling
Ecology & Evolution, 17, 571e577. preferences in fish. Animal Behaviour, 56, 611e615.
Bowling, L. C., Steane, M. S. & Tyler, P. A. 1986. The spectral distribution and Magurran, A. E., Seghers, B. H., Carvalho, G. R. & Shaw, P. W. 1992. Behavioural
attenuation of underwater irradiance in Tasmanian inland waters. Freshwater consequences of an artificial introduction of guppies (Poecilia reticulata) in
Biology, 16, 313e335. North Trinidad: evidence for the evolution of anti-predator behaviour in the
Candolin, U., Salesto, T. & Evers, M. 2007. Changed environmental conditions wild. Proceedings of the Royal Society B, 248, 117e122.
weaken sexual selection in sticklebacks. Journal of Evolutionary Biology, 20, Milinski, M. & Bakker, T. C. M. 1990. Female sticklebacks use male colouration in
233e239. mate choice and hence avoid parasitized males. Nature, 344, 330e333.
Clarke, J. M. & Schluter, D. 2011. Colour plasticity and background matching in Miller, P. J. O., Biassoni, N., Samuels, A. & Tyack, P. L. 2000. Whale songs lengthen
a threespine stickleback species pair. Biological Journal of the Linnean Society, in response to sonar. Nature, 405, 903.
102, 902e914. Mockford, E. J. & Marshall, R. C. 2009. Effects of urban noise on song and
Cummings, M. E. 2007. Sensory trade-offs predict signal divergence in surfperch. response behaviour in great tits. Proceedings of the Royal Society B, 276,
Evolution, 61, 530e545. 2979e2985.
J. L. Kelley et al. / Animal Behaviour 83 (2012) 783e791 791

Morrongiello, J. R., Bond, N. R., Crook, D. A. & Wong, B. B. M. 2010. Nuptial The influence of ontogeny and light environment on the expression of visual
coloration varies with ambient light environment in a freshwater fish. Journal of pigment opsins in the retina of the black bream, Acanthopagrus butcheri. Journal
Evolutionary Biology, 23, 2718e2725. of Experimental Biology, 211, 1495e1503.
Ord, T. J., Stamps, J. A. & Losos, J. B. 2010. Adaptation and plasticity of animal Slabbekoorn, H. & Peet, M. 2003. Ecology: birds sing at a higher pitch in urban
communication in fluctuating environments. Evolution, 64, 3134e3148. noise: great tits hit the high notes to ensure that their mating calls are heard
Peichel, C. L. 2004. Social behavior: how do fish find their shoal mate? Current above the city’s din. Nature, 424, 267.
Biology, 14, R503eR504. Stevens, M., Parraga, C. A., Cuthill, I. C., Partridge, J. C. & Troscianko, T. S. 2007.
Pitcher, T. J. & Parrish, J. K. 1993. Functions of shoaling behaviour in teleosts. In: Using digital photography to study animal coloration. Biological Journal of the
Behaviour of Teleost Fishes (Ed. by T. J. Pitcher), pp. 363e439. London: Chapman Linnean Society, 90, 211e237.
& Hall. Stuart-Fox, D., Moussalli, A. & Whiting, M. J. 2007. Natural selection on social
Ranta, E., Rita, H. & Lindstrom, K. 1993. Competition versus cooperation: success signals: signal efficacy and the evolution of chameleon display coloration.
of individuals foraging alone and in groups. American Naturalist, 142, 42e58. American Naturalist, 170, 916e930.
Reckel, F., Melzer, R. R., Parry, J. W. L. & Bowmaker, J. K. 2002. The retina of five Stuart-Fox, D., Moussalli, A. & Whiting, M. J. 2008. Predator-specific camouflage in
atherinomorph teleosts: photoreceptors, patterns and spectral sensitivities. chameleons. Biology Letters, 4, 326e329.
Brain Behavior and Evolution, 60, 249e264. Utne-Palm, A. C. 2002. Visual feeding of fish in a turbid environment: physical and
Reimchen, T. E. 1989. Loss of nuptial color in threespine sticklebacks (Gasterosteus behavioural aspects. Marine and Freshwater Behaviour and Physiology, 35,
aculeatus). Evolution, 43, 450e460. 111e128.
Rodgers, G. M., Kelley, J. L. & Morrell, L. J. 2010. Colour change and assortment in Whiteley, A. R., Gende, S. M., Gharrett, A. J. & Tallmon, D. A. 2009. Background
the western rainbowfish. Animal Behaviour, 79, 1025e1030. matching and color-change plasticity in colonizing freshwater sculpin pop-
Rosenthal, G. G. & Ryan, M. J. 2005. Assortative preferences for stripes in danios. ulations following rapid deglaciation. Evolution, 63, 1519e1529.
Animal Behaviour, 70, 1063e1070. Wong, B. B. M., Candolin, U. & Lindstrom, K. 2007. Environmental deterioration
Schultz, T. D., Anderson, C. N. & Symes, L. B. 2008. The conspicuousness of colour compromises socially enforced signals of male quality in three-spined stickle-
cues in male pond damselflies depends on ambient light and visual system. backs. American Naturalist, 170, 184e189.
Animal Behaviour, 76, 1357e1364. Wyszecki, G. & Stiles, W. S. 1982. Color Science: Concepts and Methods, Quantitative
Seehausen, O., van Alphen, J. J. M. & Witte, F. 1997. Cichlid fish diversity threat- Data and Formulae. 2nd edn. New York: J. Wiley.
ened by eutrophication that curbs sexual selection. Science, 277, 1808e1811. Young, M. J., Simmons, L. W. & Evans, J. P. 2010. Predation is associated with
Shand, J., Davies, W. L., Thomas, N., Balmer, L., Cowing, J. A., Pointer, M., variation in colour pattern, but not shape or colour reflectance, in a rainbowfish
Carvalho, L. S., Trezise, A. E. O., Collin, S. P., Beazley, L. D. & Hunt, D. M. 2008. (Melanotaenia australis). Journal of Animal Ecology, 80, 183e191.

You might also like