Morphometric data to describe phenotypical variation across Pygmy and Marbled newts (genus Triturus)

An extensive dataset on the external morphology of pygmy and marbled newts was used to explore the static allometry of sexual size dimorphism (SSD) and Rensch’s rule (RR). Females were larger in trunk and head sizes, whereas males had longer limbs and digits. Divergences in trunk and head dimensions between sexes were achieved along a common allometric slope rather than by a change in the slope’s direction, most often through a shift in the intercept. Sexual dimorphism in finger and toe length was unrelated to body size. The change in SSD with a change in body size (RR) was confirmed for female-biased traits. RR was unrelated to changes in static allometric slopes.

Methods
Five (sub)species of T. marmoratus species group were analysed: two subspecies of the marbled newts – T. marmoratus harmannis (338 females and 247 males from 46 populations) and T. m. marmoratus (80 females and 70 males from five populations) and three taxa of pygmy newts namely T. p. pygmaeus (26 males and 24 females from five populations), T. pygmaeus lusitanicus (317 females and 254 males from 38 populations) and T. rudolfi (126 females and 125 males from six populations).  A sample of the even larger crested newt T. cristatus (15 males, 30 females from one population) was included for comparison. Hybrid populations were excluded (Arntzen, 2018, 2024ab). All measured individuals were adults. The following eight morphometric traits were analyzed: snout-vent length (SVl), as representative of overall body size was measured from the tip of the snout up to and including to the insertion of hind limbs, interlimb distance (ILd), head width (Hw), head length (Hl), forelimb length (FLl), third finger length (TFl), hindlimb length (HLl) and fourth toe length (FTl). Snout-vent length as here measured purposely excludes the cloaca, the size of which is season dependent. That measure largely corresponds to the more commonly used measure up to the cloacal slit (e.g. Reinhard and Kupfer, 2015). The analysed dataset is compilation of published data (2018, 2024b, 2024c) and newly collected ones, and cover the entire distribution range (Arntzen, 2024cd).

All statistical analyses were conducted using R (R Core Team, 2022). The significance of differences in mean trait size between the sexes was tested by t-tests using the t.test function. The allometric relationships between females and males within (sub)species and among (sub)species, as well as RR were explored using SMA with the smatr package (Warton et al., 2012; 2018). First, we plotted trait measures versus SVl (all log-transformed) for each (sub)species and sex separately using sma function. The strength of the association between SVl and the specified trait was estimated by the square of Pearson’s correlation coefficient (r2). The statistical significance of the fitted allometric slope for each sex and (sub)species was estimated by the sample correlation between residuals and fitted values. We subsequently tested whether the allometric slope is non-isometric (b ≠ 1). Differentiation of the sexes was explored by testing for differences in the direction of allometric slopes (Figure 1, pattern A). When no differences were found, we tested for lateral shifts that led to a change in elevation (pattern B), shifts along the common allometric slope (pattern C), or both (pattern D).

We tested for RR using again the sma function. The log-transformed mean values for specific traits in females versus males for each (sub)species were plotted and tested for isometry (b = 1). To test for common slopes across taxa and between pairs of taxa, we used the slope.com function. First, we calculated the common slope for each of traits for all six taxa with pooled males and females as these have common slopes (Table 1) and estimated its statistical significance based on the (Bartlett-corrected) likelihood ratio statistic testing for common slope (Warton and Weber, 2002; Warton et al., 2006; Taskinen and Warton, 2013). We did post-hoc multiple comparisons of slopes by calculating the common slope for pairs of taxa for each trait. Interpretation of all statistical tests was done using the standard Bonferroni correction for multiple comparisons.

To calculate 95% confidence intervals for the values of allometric slopes in all aforementioned tests we used boot package (Davidson and Hinkley, 1997; Canty and Ripley, 2024). We defined bootstrap function and than used replicate function to create bootstrap samples. Confidence intervals were calculated by using quantile function.