Field sampling and phenotypic analysis
We noosed or captured by hand 10 males and 10 females from each population (N = 40 individuals). Captures were made in the breeding season (April-May) of 2015, and the number of ticks carried by each individual was counted in the field. The lizards were transported to the lab, where all phenotypic traits (snout-vent length, head width and length, hind leg length, number of femoral pores, and body mass) were measured, and photographs of the throat in ventral position were taken to quantify sexual coloration. For that purpose, we processed the photographs with Adobe Photoshop CS6 as explained Llanos-Garrido et al. (2017): we standardized the area of analysis using the ‘magnetic loop’ tool, we measured the red coloured surface with the ‘magic wand’ tool (with 30% tolerance) after selecting a random red-colored point, and we then used the ‘similar’ option of the magic wand, with the same tolerance, to select all areas with a similar coloration. The colored surface was measured as the percentage of colored pixels in the area of analysis. To calculate the red saturation, we use the proportion of the red channel within the RGB channel, that is: R/(R+G+B), where R, G and B are the red, green and blue channels of the graphics card. All these measures were taken blindly with respect to the population of origin. Once all this was done, the lizards were released on their site of capture.
All phenotypic analyses were performed with Statistica software (Statsoft). In order to analyze between populations differences, we used two-way ANOVAs with sex and population as factors and with different morphological measures as the dependent variables. When necessary, we used ANCOVAs to control for the effect of body size. To assess the relationship between tick load and different explanatory variables (e.g. sex, size, or number of femoral pores), analyses were restricted to the Navacerrada population. To summarize variation in male size in this population, we ran a Principal Component Analysis (PCA) that produced a single factor explaining 84.2% of the variation in the data matrix and giving high factor loadings to all size variables (snout-vent length, head length and width, mean leg length, and body mass).