Three yellow patches differently correlate with exploratory behaviour, morphological traits, leukocytes, parasites, and hormones in a lizard species
Description
Multiple within-individual non-redundant signals that convey complementary information about individual quality may have evolved in lacertid lizards (i.e., the multiple signal hypothesis ––MSH). We tested this hypothesis in a short-living species, Acanthodactylus erythrurus. We quantified the spectrophotometric reflectance of three colour patches that look yellow to the human eye and that have the typical body scheme distribution of the Lacertidae (namely cheek, lateral eyespots, and outer ventral scales) in adult males. An ultraviolet component of the three “yellow” patches is revealed by spectrophotometry. We modelled the spectral properties of these patches using model averaging and cross-validation as a function of head volume and body length (proxies of resource allocation to somatic growth), the exploratory behaviour upon a stressing situation (proxy of quality to cope with stress), body condition (proxy of nutritional state), leukocytic profiles (proxy of immune state), and faecal testosterone metabolites (proxy of reproductive state and aggression). The spectral within-individual significant differences between the three “yellow” patches together with the different relationships with the independent predictors analysed, suggested that although the three colour patches look similarly yellow to the human eye their differential expression may be influenced by patch-specific pigment allocation rules. Furthermore, the results found between the exploratory behaviour, morphological traits, leukocytic profiles, ectoparasites, and hormones with the spectral properties of the three yellow patches of A. erythrurus support that male nuptial colouration of this lizard species honestly reflect male quality and provide further support for the MSH in lacertids.
Files
Steps to reproduce
We ran all statistical models in R version 3.4.3 (R Core Team, 2017). We performed a paired test that compared the within-individual spectral reflectance of the three “yellow” patches investigated. In addition, we performed twelve models, one per spectral variable (luminance, chromaVIS, chromaUV, hue) and colour patch (cheek, first eyespot, outer ventral scale). All the models included as independent variables z-standardized Julian date of capture, body condition, head volume, body length, log10-transformed counts of both mites and blood parasites, [FTM], and the two WBC principal factors. We modelled the PC of exploratory behaviour of the lizards with the four spectral properties of the three colour patches as predictors and included [FTM] as covariate (Moore and Marler, 1987). [FTM] was later removed from this model because its effect was estimated not significant, which allowed us to include in the analysis those lizards without faecal sample (N = 41). To reduce over parametrization, we used the Akaike’s criterion to compare four separate models, each containing or luminance or chromaVIS or chromaUV or hue of the three colour patches. We checked the residual distribution and the autocorrelation of the models based on variance inflation factor (VIFs, which were < 2 for all predictors in the four models) with the vif() function of the ‘car’ package (Fox and Weisberg, 2019). We used an information criterion corrected for small sample sizes (Bedrick and Tsai, 1994) and we applied a multimodel inference approach (package ‘MuMIn’; Barton, 2018), which is a method of variable reduction recommended in ecology (Hegyi and Garamszegi, 2011). We considered sufficiently informative all models with ΔAICc ≤ 4 in relation to the best model (i.e., the one with the lowest AICc) (Burnham and Anderson, 2004). We also summed the AICc weights of all the models where the predictor appeared (i.e., conditional average) to calculate the relative importance of each variable in the averaged model. Using this procedure, we calculated the significance (α < 0.05) of the effects and their z-standardised ß coefficient ± adjusted standard error (SE). The latter z-standardization enables comparing the magnitude of the effects. The resulting final models were cross-validated using a k-fold split of 3 using the R-package ‘DAAG’ (Maindonald et al., 2015).