Introduction

In many animal species, males engage in agonistic interactions varying from threat displays to physical fighting. They do so to compete over sexual partners or over resources to attract potential mates. Physical fighting is costly, in terms of energy (Briffa and Sneddon 2007), time, and possible injury, and may increase predation risk (Glass and Huntingford 1988; Kelly and Godin 2001; Briffa and Elwood 2004). To avoid energy expenditure and potentially lethal injuries linked to fighting, many species rely on signals to assess the quality of the opponent and to mitigate the outcome of the agonistic interactions (Maynard-Smith and Harper 2003; Briffa 2014). These signals are thought to convey information on fighting ability or overall male strength or quality, often referred to as resource holding potential (RHP) (Parker 1974; Andersson 1994). Individuals may consequently assess asymmetries in RHP between themselves and opponents during intrasexual agonistic interactions using these signals (Taylor and Elwood 2003) and thus settle the interaction without escalation.

It has been suggested that intrasexual agonistic interactions may be mediated by color traits, sometimes referred to as badges of status (Rohrlich and Rubin 1975; Maynard-Smith and Harper 2003). These badges can be used to assess the RHP of the opponent if they reflect the social status or dominance of the individual (Whiting et al. 2003; Senar 2006). For example, Anolis lizards have been suggested to assess an opponent’s bite force by evaluating dewlap color, size, and display rate (Vanhooydonck et al. 2005; Steffen and Guyer 2014). Signal honesty can be maintained by an incorruptible, inherent relationship between signal and quality (i.e., relatively “cost-free”) or costs associated with a signal expression that relies on the quality of the signaller (Zahavi 1975; Searcy and Nowicki 2005; Bradbury and Vehrencamp 2011; Weaver et al. 2017). This cost could be associated with producing and maintaining signals (i.e., handicap signals, Searcy and Nowicki 2005). Producing color signals is costly as it involves pigments like carotenoids or melanin, which are also involved in the immune response (Griffith et al. 2006) and oxidative barrier (Henschen et al. 2016). Moreover, the production of melanin-based color is linked to testosterone levels which have an immunodepressive effect (Buchanan et al. 2003; Foo et al. 2017). Therefore, there appears to be a potential trade-off in the pigment allocation to badges of status versus immunocompetence and/or oxidative barriers as predicted by the handicap theory (Zahavi 1975) and the immunocompetence handicap theory (Folstad and Karter 1992). Nevertheless, the cost of ensuring signal honesty can also be from a social nature, which is mostly the case for badges of status. Here, the signal honesty is mainly guaranteed by conspecifics imposing a high cost on dishonest individuals by punishing those individuals (i.e., significantly more aggressive behavior toward cheaters) (Rohwer 1982; Tibbetts and Dale 2004; Tibbetts and Izzo 2010).

Although pigment-based colors have been extensively studied in this context, structural colors are clearly also important, notably ultraviolet (UV) coloration. Over the past decade, a growing body of evidence has been accumulated on the involvement of UV coloration in intrasexual competition and the role of these UV color traits as badges of status (Lim and Li 2013; Xu and Fincke 2015; Martin et al. 2016; Sabol et al. 2017). A higher UV reflectance appears to convey information on RHP (Stapley and Whiting 2006; Whiting et al. 2006; Tringali and Bowman 2012; Lim and Li 2013), a better immune response (Doucet and Montgomerie 2003; Griggio et al. 2010; Megía-Palma et al. 2016), or a higher testosterone level (Roberts et al. 2009). The signal honesty of UV signals has been suggested to be ensured by the immunocompetence handicap mechanism (Folstad and Karter 1992; Roberts et al. 2009) but mainly through a social cost (Martin et al. 2016).

Interestingly, some animals can exhibit rapid color change (in a few minutes or even less than a second), which they may use in the course of intraspecific disputes to communicate (Hutton et al. 2015). This change is found in a wide range of taxa: for example, cephalopods (Adamo et al. 2000), lizards (Korzan et al. 2006; Batabyal and Thaker 2017), and insects (Umbers et al. 2013). Nevertheless, most studies on the use of color change during social interactions have investigated this ability as a simple ON/OFF feature, without considering the dynamic aspect of color change (i.e., speed, temporal variation, direction). Interestingly, the dynamics of color change likely offer as many communication possibilities and complexity as vocalizations. For example, a previous study by Ligon and McGraw (2013) investigated the role of dynamic color change in intrasexual agonistic interactions in chameleons. They demonstrated not only that chameleons exhibit submissive behavior through the exhibition of darker colors (Ligon 2014) and that brighter colors act as badges of status (Ligon and McGraw 2016), but also that the speed by which the color change occurs matters (Ligon and McGraw 2013). They further suggested that the honesty of this signal is guaranteed by a social cost (Ligon and McGraw 2016). However, the UV component of color change has been neglected even though numerous color-changing animals are sensitive to UV (Briscoe and Chittka 2001; Bowmaker 2008), including chameleons (Bowmaker et al. 2005). During intrasexual agonistic interactions, these animals likely show color changes in the visible spectrum and the UV range. This realization raises the question of whether they rely on color change in the visible range only or on both the visible and UV range to communicate and settle social interactions.

Chameleons (Squamata: Chameleonidae) are an excellent biological model to investigate the role of dynamic color change in intrasexual agonistic interactions in both the visible and UV ranges. Selection for conspicuous signals is likely to have driven the evolution of this group’s color displays (Stuart-Fox and Moussalli 2008). Moreover, they are highly territorial animals that often engage in physical and sometimes even deadly fights (Tolley and Herrel 2013). Furthermore, some species exhibit complex color changes (Nečas 1999; Tolley and Herrel 2013; Teyssier et al. 2015) including in hue, brightness, and saturation, offering a large color-change repertoire. Finally, chameleons are tetrachromatic lizards with UV-sensitive photoreceptors (λmax = 375–385 nm; Bowmaker et al. 2005).

In the present study, we investigate the role of color change in the visible and the ultraviolet range during intrasexual agonistic interactions in panther chameleons (Furcifer pardalis, Cuvier 1829). To do so, we staged dyadic agonistic encounters among male F. pardalis under appropriate light conditions (i.e., including UV) and examined the dynamic color change during those interactions. We predict that, as in C. calyptratus (Ligon and McGraw 2013; Ligon 2014), male F. pardalis will rely on brightness changes to settle agonistic interactions. Moreover, we predict that ultraviolet signals may also be used as chameleons perceive light in the UV range. Finally, we tested whether saturation and hue changes are involved in settling agonistic interactions among males.

Material and methods

Animals and husbandry

Furcifer pardalis is a diurnal and arboreal lizard endemic to Madagascar and found in a wide range of habitats along the northern and eastern coasts of Madagascar. This species exhibits a strong sexual dimorphism and considerable intraspecific variation in male coloration: females and juveniles are tan to brown with hints of pink or orange, while adult males are much larger and have various combinations of bright red, green, blue, and yellow. This polychromatism among males depends on the region of origin and is either termed a “morph” or “locality.” Local variation also appears to exist within morphs (Ferguson 2004). For this study, nineteen adult (over 6 months old) captive-bred males Furcifer pardalis of the “Ambilobe” morph were used.

Animals were kept individually and were visually isolated from one another in mesh terraria (46 × 46 × 91 cm, ReptiBreeze, ZooMed) outfitted with branches and plastic plants to provide hiding spots. Animals were kept in a dedicated room at the Paris Zoo (Parc Zoologique de Paris). The room temperature was maintained at 26 °C during the day, and the temperature dropped to 22 °C at night. Fluorescent tubes providing 12% UVB (Reptile Lamp 12% T8, Arcadia) and a 40 W heating bulb (Repti Basking Spot, ZooMed) were suspended above each cage. The photoperiod was set at 12/12 h. The animals were fed thrice weekly, and crickets were calcium dusted once a week. Water was provided to the animals during three daily misting periods (9 am, 12 pm, and 4 pm) with an automated misting system (Vivaria project) and 20cL drippers.

Experimental design

We used a large arena (144 × 50 × 80 cm) with opaque Plexiglas sides and a front made of transparent Plexiglas of 50 cm high to allow for behavioral observations and photo/video recordings. As chameleons are perch-dwelling lizards, we provided artificial branches to simulate an arboreal environment. The overall setup was illuminated with a combination of nine light sources placed 56 cm above the setup as described previously (Dollion et al. 2020). Combining the different light sources enabled us to cover most of the solar spectrum, including infrared with the halogen bulbs, UVs with the fluorescent tubes, and visible light with the other sources (Fig. S1). Hence, this combination provided a light environment in accordance with the chameleon visual system, which includes UVs in the UVA range (Bowmaker et al. 2005). In the middle of the arena, a thin (~ 0. 7 mm) transparent PVC plate (50 cm × 100 cm), hereafter called barrier, was placed to prevent direct attacks and injuries while allowing the animals to see one another (Fig. 1).

Fig. 1
figure 1

source positions from the top: 1 = 100 Watt. 2800 Kelvin. 1320 lm halogen bulbs, 2 = 60 Watt. 2700 Kelvin incandescent bulbs, 3 = 60 Watt. 4000 Kelvin. 806 lm LED bulbs, 4 = UVB fluorescent tube ReptiSun® 10.0 High Output UVB Bulb, 5 = Reptile systems New Dawn T5 LED, and 6 = Arcadia T5 D3 + Desert 12% Reptile Fluorescent Lamp (figure adapted from Dollion et al. 2020)

Picture and schematic representation of the experimental setup. a Picture of the setup from one of our cameras. b Schematic drawing of the setup showing where the cameras were positioned. c Schematic drawing of the light

Full size image

From April 2018 to May 2018, round-robin tournaments were performed to assess the dominance status between our 19 male panther chameleons. Males were split into four groups of four to five different males balanced in size and color pattern, meaning that individuals with similar color patterns (i.e., blue bar and red bar) or size were dispatched in different groups. The round-robin was performed within groups, resulting in three to four encounters per male, totalling 42 agonistic interactions. The encounters occurred at room temperature (26 °C) from 10 am to 6 pm, corresponding to their daily activity. The arena was sprayed and cleaned with clear water before each trial to prevent potential effects of odors remaining from a previous trial. Animals were able to interact for 30 min unless one of the animals crossed the barrier upon which the trial was halted to avoid physical injury to the animals. The behavior was recorded with an HD camera, HDCR-CX740VE (SONY, Minato-ku, Tokyo, Japan). It was impossible to record data blindly because our study involved focal animals, yet we randomized the order and time of the encounters. Agonistic interaction outcomes were assessed through behavioral observations (Table 1). Animals were considered as “winners” if they succeeded or tried to cross the barrier, if they chased (followed) their opponent from their side, or if they spent most of the interaction trying to climb the barrier (Table 1). Animals were considered as “losers” if they fled from their opponent while trying to maintain a considerable distance between them. In some cases, individuals exhibited “winner”-like behavior at the beginning of the interaction, but then the direction (winner/loser) of the interaction changed. In this case, we only considered the final behavior. We also observed some “ties” where both animals appeared to ignore each other or were facing and trying to fight each other (hitting the barrier violently in turn) until the end of the interaction (Table 1). We chose not to consider these “tie” situations for our statistical analysis, as, in nature, one of the individuals would likely have prevailed.

Table 1 Ethogram used to determine the outcome of the agonistic interactions
Full size table

Color calibration and measurements

During the disputes, pictures of both protagonists were taken twice every 2 min with a full spectrum converted camera, Samsung NX-1000. One picture was taken in the visible range (VIS) and a second in the ultraviolet range (UV) immediately after the first. For pictures in the visible range, a filter blocking ultraviolet and infrared was manually placed in front of the camera (UV/IR cut/L Filter, Baader ©, Mammendorf, Germany), and pictures were taken with a 1/640-s exposure. Immediately after the picture was taken, the filter was changed for a filter blocking all wavelengths except those ranging from 320 to 400 nm (Venus-U Planetary Filter, Optolong ©, Kunming City, China), and a picture was taken with a 1-s exposure.

For the color calibration in the visual range, a color checker (SpyderCHECKR ™) was photographed while placed at thirteen different regions of the arena, once empty. Color calibration was performed using Adobe Photoshop Lightroom 6 and the SpyderCHECKR software (version 1. 2. 2) provided with the SpyderCHECKR ™, following the instructions. As each of the two opponents might be in different regions of the arena, images were cropped to isolate each opponent. Then, according to the position of the individuals in the arena, the corresponding color calibration was applied. Calibrated images were then used for color measurements. As the UV filter imposes a narrow hue range resulting in a constant pink coloration, color calibration was not required for the UV pictures.

Color measurements were performed by retrieving RGB values using the RGB measure tool in ImageJ. We measured fifteen squares of sixteen pixels (NVIS = 8; NUV = 7; Fig. 2), describing the specific color patterns of male F. pardalis. In the visual spectrum, we quantified color at the bands (N = 3), interbands (N = 3), and the lateral line (N = 2); in the UV, areas were defined as absorbing UV (N = 5) (i.e., bands, eyelid, and head bony tubercles) or reflecting UV (N = 1) (i.e., lips) and the lateral line (N = 1) (Fig. 2). The lateral line was considered independently because it was either absorbing or reflecting UV according to the individual and time. RGB values were then compiled in R (R Core Team 2019) and converted into HSV values (H, hue; S, saturation; V, brightness) using the Colorscience package (Gamma and Davis 2018). HSV is an alternative representation of the RGB color model that aligns with humans’ color-making attributes and color perception. As pictures were taken every 2 min, each picture corresponds to a time step of two minutes, with the first picture of each interaction corresponding to t = 0 min.

Fig. 2
figure 2

adapted from Dollion et al. 2020)

Color patterns sampled during male-male F. pardalis agonistic interactions. As interindividual variation exists. we focused on common patterns. Body regions were characterized as bands (○). interbands (□), and lateral line (Δ) in the visible spectrum (VIS; 400 to 700 nm). However. for the UV range (UV: 320 to 400 nm), body regions were different and characterized by their UV properties as absorbing (○), reflecting (□), and lateral line (Δ) which either absorb or reflect through time and according to the individual. (Illustration by Julien NORWOOD) (figure

Full size image

From our measurements, Euclidian distances between colors over two time steps (1 and 2) in the 3D HSV color space were calculated (DEHSV) (Eq. 1). The following six derived variables, which allow us to describe the color changes occurring during social interactions, were computed for each color value (i.e., hue, saturation, brightness, and DEHSV) at each of the fifteen squares selected on the body of the chameleon: the variance, the maximum speed of change between two time steps, the maximum absolute color change (Eq. 2), the overall absolute color change (Eq. 3), the maximum color variation (Eq. 4), and the overall color variation (Eq. 5). DEHSV values enable us to the changes occurring in the three dimensions of the HSV color space in one measurement. All variables (Fig. 3; Dollion et al.2020) were calculated independently for UV and VIS pictures and averaged by body region (VIS, bands, interbands and lateral line; UV, absorbing, reflecting and lateral line). Maximum UV brightness and maximum UV saturation were also retrieved.

Fig. 3
figure 3

Schematic representation of the different computed color change variables. Where X represents any color value (e.g., hue. saturation. brightness, or DEHSV). a Overall color variation in light blue and overall absolute color change in dark blue. b Maximum color variation in light red and maximum absolute color change dark red

Full size image

In our formulae, x represents a color value (H, S, V or DEHSV); tmin represents the beginning of the interaction; tmax represents the end of the interaction; txmax is the time at which x reaches its maximum value; txmin is the time at which x reaches its minimum value; xtmax is the x value at the end of the interaction; and xtmin is the x value at the beginning of the interaction. In Eq. 1, H is the hue value, S is the saturation value, and V is the brightness value.

$${DEHSV}_{x1-x2}=\sqrt{{\left({S}_{1}\mathrm{cos}{H}_{1}-{S}_{2}\mathrm{cos}{H}_{2}\right)}^{2}+{\left({S}_{1}\mathrm{sin}{H}_{1}-{S}_{2}{\mathrm{sinH}}_{2}\right)}^{2}+{\left({V}_{1}-{V}_{2}\right)}^{2}}$$
(1)
$$\frac{{X}_{max}-{X}_{min}}{\left|{t}_{Xmax}-{t}_{Xmin}\right|}$$
(2)
$$\frac{\left|{X}_{tmax}-{X}_{tmin}\right|}{{t}_{max}-{t}_{min}}$$
(3)
$$\frac{{\sum }_{t={t}_{Xmin}}^{{t}_{Xmax}}\left(\left|{X}_{t+1}-{X}_{t}\right|\right)}{\left|{t}_{Xmax}-{t}_{Xmin}\right|}$$
(4)
$$\frac{{\sum }_{t={t}_{min}}^{{t}_{max}}\left(\left|{X}_{t+1}-{X}_{t}\right|\right)}{{t}_{max}-{t}_{min}}$$
(5)

We were unable to model the color vision of the panther chameleon because the raw data required to calculate the cone-catch values for visual system modeling (Vorobyev and Osorio 1998; Siddiqi 2004; Troscianko and Stevens 2015) are no longer available for Furcifer pardalis (Bowmaker, personal communication).

Statistical analyses

Before analyses, the distribution of each color-change variable was transformed where needed using a Box-Cox power function with “AID” package to meet the requirements of a normal distribution (Asar et al. 2017).

We summarized color-change information using principal component analysis(PCA) on the centered and scaled individual values of each color change variable with the “ade4” package (Dray and Dufour 2007). For the visible range (VIS), all variables were incorporated in the PCA, while for the UV range (UV), only brightness and saturation were used, as the hue was constrained by the filter and nearly constant. The number of principal components (PCs) used for the subsequent analyses was chosen using the broken stick method (Legendre and Legendre 1998) while representing at least 70% of the variability and for which the contribution of each variable to PCs provided a relevant interpretation of the color change.

To determine whether the color change exhibited by an individual could explain the outcome of an intrasexual agonistic interaction, generalized linear mixed-effects models (GLMM) (“glmer”) using “lme4” package (Bates et al. 2015) were performed. We tested for the effect of color change (i.e., PCs and maximum brightness and saturation for the UV range) either in the visible range or in the UV range, on the agonistic interaction outcome, considering the male identity and the test order as random factors. These were tested for each body region separately because we found a significant effect of the body region on most of the color change variables (P < 0. 01)in both VIS and UV (see Table S1). Male size (i.e., snout-vent-length) was not incorporated in any model as there was no significant effect on the agonistic interaction outcome (GLMM: Χ2 = 3. 77; P = 0. 052).

We evaluated the relative importance (RI) of each predictor variable using model-averaging approaches (Burnham et al. 2011) within model sets for each body region from each spectral range to build one model per region. From the model averaging results for each region, we kept the variables with an RI value exceeding 50%.

All statistical analyses were conducted in R version 3.6.1 (R Core Team 2019); model selection and model-averaging were also undertaken using the R package MuMIn (Barton 2019).

Results

PCA results and interpretation

The first three principal components accounted for 70.47% of the total variance in the visible spectrum. The first principal component (PC1) of our PCA describes hue and saturation changes, and the second principal component (PC2) depicts brightness and DEHSV changes (Table 2). The third principal component (PC3) describes brightness changes (Table 2). In the UV, the first two principal components accounted for 73.93% of the total variability. The first principal component (PC1) represents UV color changes, whereas the second principal (PC2) describes UV brightness changes (Table 2).

Table 2 Contributions of original variables to principal components (PC) retained for further analysis. Components accounted for at least 70% of the total variability. In the visible range (VIS), 3 PCs were kept and accounted for 70.46% of the total variability. In the ultraviolet range (UV), 2 PCs were kept and accounted for 73.94% of the total variability. Bold values represent high loadings. Variables were computed from the images of the 42 interactions among male Furcifer pardalis. In each image, the color was measured from the band, interband, and lateral lines in the range of 400–700 nm and from UV-absorbing region, UV-reflecting region, and lateral line in the UV range (320–400 nm)
Full size table

Male color change and agonistic interaction outcome

In the visual range, multi-model averaging uncovered PC2 (i.e., brightness and DEHSV changes) as being the best predictor for the band models (RI PC2 = 95%) and lateral line models (RI PC2 = 99%; Fig. 4D). Similarly, multi-model averaging on the interbands models uncovered PC1 (i.e., hue and saturation changes) (RI = 65%) and PC2 (RI = 99%; Fig. 3D) as important predictors. GLMM on those selected models show that males displaying more brightness and DESHV changes (PC2) at the bands (Fig. 4a; Fig. 5a) and the interbands (Fig. 4b; Fig. 5b) are significantly more likely to win. Although a tendency could also be observed for the lateral line, GLMM results are not significant for this body region (Fig. 4c). For the interbands, GLMM results did not show a significant effect of the hue and saturation changes (PC1) on the probability of winning (Fig. 4b). In the UV range, multi-model averaging did not uncover any of our UV variables to be good predictors of the winning probability (i.e., RI < 50%; Fig. 4d; Fig. 6).

Fig. 4
figure 4

Relationships between winning probability and color change variables in the visible range (VIS) selected by the model selection by AICc and the relative importance of color variables to models, based on multi-model averaging, in both visible and UV range. Ribbons illustrate the 95% confidence interval. A winning probability of 100% corresponds to “winner,” while zero corresponds to “loser”. a Relationship between winning probability and color change at the bands. b Relationship between winning probability and color change at the interbands. c Relationship between winning probability and color change at the lateral line and d RI values of color change variables (PCs) in the visible range predicting intrasexual agonistic interaction outcome

Full size image
Fig. 5
figure 5

Effect of the color change in the visible range in adult males F. pardalis on the outcomes of intrasexual agonistic interaction. Boxplots showing respectively: a Color changes occurring at the bands. b Color changes occurring at the interbands. c Color changes occurring at the lateral line. PC1 = hue and saturation changes, PC2 = brightness and DEHSV changes, PC3 = brightness changes

Full size image
Fig. 6
figure 6

Effect of the color change in the UV range in adult males F. pardalis on the outcomes of intrasexual agonistic interaction. Boxplots showing respectively: a Color changes occurring in regions that absorb UVs. b Color changes occurring in regions that reflect UVs. c Color changes occurring at the lateral line. PC1 = UV-color changes. PC2 = UV-brightness changes

Full size image

Discussion

This study highlighted that color change in the visible range plays a key role in agonistic interaction outcomes in male F. pardalis. We showed that animals that exhibited more brightness and DEHSV changes (PC2) were more likely to win a contest. These results suggest that brightness changes play a prominent role in agonistic interactions in F. pardalis. This is likely the case for chameleons in general as our results align with previously published data for the veiled chameleon (Chamaeleo calyptratus, Duméril and Duméril 1851) (Ligon and McGraw 2013). Similar to our results, the dynamics of the brightness changes matter in C. calyptratus (Ligon and McGraw 2013). In this species, males that brighten faster were more likely to win a fight. Furthermore, brightness changes affected the likelihood of winning, but only at specific body regions, namely the bands and interbands. Hence, those two regions seem to have significant involvement in communication during agonistic interactions. Contrary to bands and interbands, the lateral line color change did not significantly explain the agonistic interaction outcome in male F. pardalis. Therefore, the lateral line appears not to be involved in intraspecific communication. However, it likely plays a role in antipredator defense as a disruptive coloration pattern (Stevens and Merilaita 2009). To better understand the importance of patterns in communication in chameleons, further studies would benefit from quantifying those patterns using quantitative color pattern analysis (QPCA) (van den Berg et al. 2020) and doing so across time to investigate pattern changes as well.

Our results highlight the dynamic properties of color change and pinpoint the importance of the frequency of color change (i.e., maximum color variations and overall color variations; see Fig. 3). In other words, the number of times an individual changes color is a, important factor. Consequently, what we might observe here is that winners exhibit several brightness changes while losers exhibit a few or only one, namely darkening. During our experiments, we noticed that losers were generally darker than the winners at the end of the interaction. As in C. calyptratus (Ligon 2014), we can hypothesize that male F. pardalis darken to communicate submission. Submissive behaviors are quite common (Lorenz 1966) and should be favored by natural selection because they would avoid unnecessary expenditure of time and energy on both sides (Matsumura and Hayden 2006). Brightness changes in chameleons are also involved in the context of camouflage in some species of Bradypodion, which decrease their brightness when exposed to highly visual predators (Stuart-Fox et al. 2008). Consequently, we might predict chameleons to increase their brightness to increase conspicuousness to communicate or to decrease it to prevent injuries from conspecifics and from being spotted by predators.

Unexpectedly, we found no relationship between UV coloration nor its dynamics and the outcome of intrasexual agonistic interactions, despite its presence in male F. pardalis (Fig. 1). This finding suggests that UV signals are likely not involved in chameleon communication during intrasexual competition, contrary to color changes in the visible range. This might be explained by the fact that our illumination spectrum (Fig. S1) presents a gap (low number of photons) between 360 and 420 nm, which corresponds to the peak sensitivity for the chameleon photoreceptors (375–385 nm; Bowmaker et al. 2005). Hence, our individuals may use UV signals, yet our illumination spectrum may not be bright enough for our individuals to detect those. The use of UV within this specific range would be beneficial for this chameleon because the spectral range around 430 nm also seems to correspond to a spectral sensitivity gap in avian vision (Hart and Hunt, 2007; Hart and Vorobyev, 2005). Therefore, chameleons might use UV signals within this specific range as private signals, but we could not assess this with our experimental design. Even though UV color change might not be involved in agonistic integrations, UV coloration and UV color change may still be involved in chameleon communication in other kinds of social interactions than agonistic. Investigating agonistic intrasexual interactions among males to study the function of rapid color change is of interest and has been studied in detail (Adamo and Hanlon 1996; O’Connor et al. 1999; Korzan et al. 2006; Umbers et al. 2013; Ligon and McGraw 2013; Batabyal and Thaker 2017). However, the role of color change in other contexts like mate choice remains poorly investigated. Yet, a recent study (Dollion et al. 2020) showed that color change is involved in chameleon mate choice and that females seem to rely on a color change in both the visible and UV range. Consequently, our understanding of chameleon communication, and more broadly the communication function of rapid color change, will benefit from future studies investigating this phenomenon across diverse social contexts.

Our study would have benefited from using visual modeling (Vorobyev and Osorio 1998; Siddiqi 2004; Troscianko and Stevens 2015) as this would have provided an animal vision perspective of the observed color changes. Unfortunately, the raw data required for such analyses were not available at the time of our study (Bowmaker, personal communication). Although the spectral sensitivities of other chameleons, like Chameleo dilepis, could have been used, the study of Bowmaker and colleagues (2005) showed that the spectral sensitivity of chameleons might vary greatly between genera. Future studies would benefit from measuring spectral sensitivity in Furcifer pardalis. These data could then be used to define color patterns based on the spectral sensitivity of F. pardalis using JND (just noticeable difference) approach (Pike 2012), similarly to Ligon and McGraw (2013) but here taking into account the UV range as well. Moreover, given the subspecies diversity and strong dichromatism in Furcifer pardalis, it would be of interest to investigate whether the spectral sensitivity differs between subspecies and/or between males and females.

The cost of different physiological pathways associated with different components of color change may be different. For example, changes in brightness are the result of the dispersion of melanosomes within dermal melanophores. The process of melanosome translocation is assumed to be achieved by the microtubules and actin filaments of the cytoskeleton and the associated motors, kinesin and dynein (Ligon and McCartney 2016). However, these motors consume energy (Hackney 1996), suggesting that brightness changes may be costly. Although the mechanisms through which melanophores and xantho-erythrophores allow color change are very similar, the function of iridophores is quite different. Iridophores enable color change thanks to the modification of the space between photonic nanostructures, here guanine platelets, within the iridophores (Teyssier et al. 2015; Ligon and McCartney 2016). Iridophores are thought to be the principal actors of color change in the panther chameleon, in which small changes in the space between guanine platelets lead to noticeable differences in color (i.e., hue) (Teyssier et al. 2015). Changing the arrangement of guanine platelets might thus be cheaper than manipulating brightness through melanosome translocation. If so, winners might invest more energy in changing brightness, yet the actual cost of color change remains unknown to date.

To test these ideas, future work should focus on the physiological cost associated with each component of color change. This work would provide insight into the information content of this dynamic signal which remains poorly investigated. Along the same vein, color change involves several types of pigments, including carotenoids in xantho-erythrophores and melanins in melanophores, pigments known to play key roles in immunoregulation, immunostimulation, lymphocyte proliferation, and free radical scavenging (Moller et al. 2000; Galván and Solano 2009). Thus, the differences in color-change ability observed in this study may be related to high quantities of circulating pigments in dominant individuals. These individuals would then also be able to allocate these pigments to both immune system functioning and antioxidant protection in addition to their use in color signals. It would also be interesting to explore this for UVs as UV coloration can also reflect individual quality, including immunocompetence (Martín and López 2009; Griggio et al. 2010) and testosterone levels (Roberts et al. 2009).

To conclude, this study underlines the importance of brightness changes in the visual spectrum to settle agonistic interactions in male F. pardalis. Unexpectedly, UV seems not to be involved in male intrasexual interactions in this species.