, Ixodidae) was counted for each individual (ticks are easily detected on the body surface). Colour variables of the throat were measured with an Ocean Optics USB4000 spectrometer, using a DT-Mini-2-GS

light-source and a QR400-7-SR/BX reflection probe, single end fixed in an RHP1 holder (Ocean Optics Inc, Dunedin, FL, USA), explained in detail earlier by Bajer et al. (2010, 2011). Briefly, three independent measurements on different, randomly chosen spots of the throat were recorded for every lizard, using a separate probe contact per measurement, and the average buy A-769662 was calculated for each individual. Throat reflectance was characterized by total brightness (R320–700), UV chroma (R320–400/R320–700) and blue chroma (R400–490/R320–700) (Whiting et al., 2006). Principal components analysis was performed on the three head variables. The first principal component (Head PC) described 90% of the total variation (eigenvalue = 2.69), and showed positive correlation with all original variables (factor loadings: head height = 0.94; head length = 0.95; head width = 0.96). The Head

PC scores were used in the subsequent statistical analyses. The number of ectoparasites were log10 transformed (Log10Par) for better distribution. We used general linear models (GLMs) to test for correlations between different throat colour traits (UV chroma, blue chroma, total brightness) and other individual Angiogenesis inhibitor characteristics. We are aware of the problem imposed by the non-independence of these colour variables, but because both UV and blue chroma are calculated from brightness, we decided to analyse them separately. Each GLM was run with identical predictor variables (SVL, BW, Head PC, TL, FP, DA and Log10Par) and year of capture as random factor. We applied backward stepwise model selection. Non-significant explanatory variables were deleted one by one in decreasing

order of P, and final models included only the significant main effects. SVL and year of capture was never removed from the models in order to keep them for correction. Model selection based on all the P-value is considered conservative in comparison with, for example, the selection methods based on Akaike’s or Bayesian information criteria, and differs very little from the others in its predictive ability (Murtaugh, 2009). DA in all these models was represented as the signed differences between right and left femoral pore numbers. However, because of the problem of separating directional and fluctuating asymmetry and the information content of directional asymmetry (see above), we also ran these models with the absolute difference between sides. Whenever the results differed qualitatively, we report them in addition to the original models. All analyses were performed using the SPSS 17 (SPSS Inc., Chicago, IL, USA) software.