Computational protocol: The brown anole dewlap revisited: do predation pressure, sexual selection, and species recognition shape among-population signal diversity?

Similar protocols

Protocol publication

[…] We measured dewlap reflectance at the centre of the dewlap, using an Avantes spectrometer (AvaSpec-2048 USB2-UA-50; Avantes, Apeldoorn, the Netherlands, range 250–1000 nm) and deuterium-halogen light source (AvaLight-DHS; Avantes, Apeldoorn, the Netherlands) equipped with a fibre-optic probe. The probe was mounted within a metal holder to ensure readings at a fixed distance from the surface and was held perpendicular to the surface of the maximally extended dewlap. Reflectance data were collected for wavelengths from 300 to 700 nm, including the lower range of photon absorption by UV-sensitive photoreceptor cones published for anoles (). To investigate dewlap colour variation, we interpolated each spectrum to 1 nm wavelength intervals and extracted four variables following : brightness, hue, and relative reflectance in UV (RF 365 nm) and in red (RF 655 nm) (). We calculated brightness as the total area under the uncorrected spectral curve (300–700 nm) (; ). For the remaining three colour variables, we corrected each spectrum for brightness by making the area under the curve equal to 1 (). Hue was defined as the cut-on wavelength, i.e. the midpoint between baseline and maximum reflectance (; ; ; ). We decided to extract relative reflectance specifically in UV (365 nm) and red (655 nm), as the A. sagrei dewlap spectrum shows maxima and a high level of intraspecific variation at both wavelengths (; T. Driessens, 2017, personal observation). Spectral measurements were carried out for 242 males and 217 females in total, distributed across nine populations. We do not have spectral data for the seven population sampled by , and for the population from Central Cuba (Santa Clara), due to technical problems with the spectrometer. All analyses of spectral data were run in R using the ‘pavo’ package (). [...] In this study, we considered interpopulational variation and therefore used population means and proportions of individuals per population as data points. We applied arcsine square root transformations to all proportion data (dewlap pattern, TBF, model attack rate) to meet normality assumptions ().Our statistical analyses differed from the ones adopted by in two important ways. Firstly, adopted an information-theoretic approach to compare seven plausible models of selection for dewlap size and pattern. In doing so, they tested which combination of multiple predictor variables best describes the variation in dewlap characteristics. We are reluctant to take this approach because of the limited number of data points (seven populations in their case, nine to 17 in our extended dataset) relative to the number of predictors (see also ; ). Rather, we opted for univariate regression analyses, linking individual predictor variables to individual dewlap characteristics. This allowed us to make full use of the information available for a particular pair of predictor and response variable (not all variables could be measured in all populations). Second, in analysing the data here, we took phylogenetic relationships among the study populations into account. The phylogeny used in the comparative analyses is the one proposed by , which is based on mtDNA haplotypes obtained by . Phylogenetic regression analyses were conducted using the pgls() function in the ‘caper’ package in R (; ). This method uses maximum likelihood to simultaneously estimate the regression model and the phylogenetic signal (Pagel’s λ) of the residual error (; ). It has been shown to do better than a priori tests of phylogenetic signal to estimate the appropriateness of phylogenetically corrected tests, especially when sample sizes are smaller than 20 (; ; ). Comparisons of dewlap characteristics between islands with and without Leiocephalus lizards were conducted using the phylANOVA() function in the ‘phytools’ package in R (). Because data from one population (San Salvador) could only be obtained outside the breeding season (see Materials and Methods, section ‘Animals’), we have run all analyses with and without inclusion of this population. Results were nearly identical and we will, therefore, report results for the complete dataset only. Raw P-values were corrected for multiple testing using the Benjamini–Hochberg (BH) procedure (). All statistical analyses were conducted in R version 3.2.1 (). […]

Pipeline specifications

Software tools pavo, PHYSIG, Phytools
Applications Miscellaneous, Phylogenetics
Organisms Homo sapiens