Results

Shape and size

The first two principal component (PC) axes of humerus shape explained 34.5% (PC1) and 13.7% (PC2) of the observed shape variation (Fig.2).Echinotriton andersoni occupied a slightly different morphospace than Tylototriton , although there was some overlap especially with T. uyenoi and T. kweichowensis . Generally shape changes on the first PC corresponded to humerus thickness, representing thick humeri for negative PC1-scores and thin humeri for positive PC1-scores. The second PC corresponded to the thickness of the middle part of the humerus and the orientation of the Crista dorsalis humeri. According to that, Echinotriton exhibits more robust humerus shape compared to Tylototriton , T. taliangensis exhibiting the most gracile humerus. PC1 explained 20.1% and PC2 11.6% of the observed shape variation (Fig.3) of the PCA of cranial shape. Echinotriton was well separated from all other species of the genusTylototriton , but T. asperrimus showed the highest similarity to Echinotriton . Among Tylototritonspp., T. kweichowensis occupied the most distinct morphospace (Fig.3). Echinotriton -skull shape (positive PC1 scores) showed a robust skull with a strong maxillary connection to the quadrate and pterygoid. The snout was more pointed and with lower and anteriorly ranging nasals than in all Tylototriton . The second PC axis corresponds to the height of the fronto-squamosal arch in relation to the skull roof, the posterior extent of the squamosal and the occiput width.
Procrustes ANOVA revealed a strong allometric effect both in humerus and cranial shape (Tab.1). The allometric trajectories differed interspecifically (Fig.4) but not intersexually (Fig.5), indicated by a significant interaction of size and species but not size and sex (Tab.1). In allometric trajectories of the humerus, T. uyenoi andT. shanorum exhibited a different direction (Fig.4a) while in cranial shape, especially T. asperrimus showed a different pattern (Fig.4b). Accounting for sexual allometry only, generally large specimens exhibited a thinner humerus (Fig.5a) whereas in the cranium the dorso-lateral ridge became more elaborate and the connection of the maxillary with the pterygoid and quadrate turned more pronounced and the quadrate shifted more posteriorly (Fig.5b).

Sexual dimorphism

The Procrustes ANOVA revealed different interspecific patterns of sexual dimorphism, indicated by the significant interaction of species and sex (Tab.1). In order to test our hypothesis that the mating mode might explain different shapes, we carried out additional Procrustes ANOVAs on shape including mating mode as explanatory variable. The mating mode explained a significant amount of variation, with and without log(CS) as covariate, in the humerus and cranium (Tab.2). In general the species applying a circle dance had more robust and thicker humeri including an elaborated crista dorsalis and a higher crista ventralis compared to species employing a ventral amplexus (Fig.6). The cranium of circle dancers was wider at its occiput, exhibited a shorter frontal arch, less connection between the maxillary bone and the quadrate and pterygoid, a higher snout tip, longer vomerine tooth rows and more distal internal nares. The interaction term of mating mode with sex was significant in cranial shape when accounting for size as covariate (Tab.2 b), indicating different SShD-trajectories between the two different mating modes among identical size classes (Tab.2 b, d). We found no indication for different SShD-trajectories between mating modes in humerus shape.
As the analysis indicated different SD-trajectories, we performed a trajectory analysis with sex as grouping factor to figure out and illustrate shape changes in different trajectories across male and females between species (Fig.7, 8). Echinotriton andersoni did not differ markedly from Tylototriton spp. In the latter, cranial SD-trajectories showed contrary directions between some species (Fig.8). Pairwise species-comparison revealed only one pair of species (T. asperrimus : T. taliangensis ) with an alpha-level below 5% and six species pairs below 10% for humerus SShD patterns and two species pairs below 5% and 10% (T.himalayanus : T. kweichowensisand T. shanjing : T. verrucosus ), respectively in SShD patterns of cranium shape (Appendix Tab. A2). To illustrate the different male to female SD-trajectories we plotted TPS-grids for amplecting T. himalayanus and circle dancing T. kweichowensis , both species deviating strongly in SD-trajectories (Fig.7, 8). In T. himalayanus, the humerus turned thinner in the middle part while the distal end was more twisted in females. Further, the crista ventralis was slightly more pronounced in females. InT. kweichowensis especially the crista dorsalis appeared more elaborate in females (Fig.7). Male to female shape changes in cranium morphology of T. himalayanus included an elaborated squamosal bony ridge, a posterior shift of the quadrate, a stronger connection of the maxillary with quadrate and pterygoid, lower nostrils and a shorter frontal arch. In T. kweichowensis cranial shape changes between sexes were much less pronounced and comprised a posterior shift of the quadrate and a slightly posteriorly shift of the palatal fissure between the vomers. Shape changes from the mean shape to male and female shape, respectively, were similar in the humerus, but differed between males and females in their extent, whereas cranial shape changes to the mean deviated between sexes of T. himalayanus but not in T. kweichowensis (Fig.9).
Procrustes ANOVA on humerus and cranium log(CS) of species and sex revealed interspecific but also intersexual differences in size (Tab.3 a, b). Further, SSD differed between species indicated by a significant interaction of species and sex (Tab.3 a, b) . Analysis of the effect of mating mode on size yielded no general size differences in the humerus between dancing and amplecting species but the interaction of sex and mating mode was close to significance level (Tab.3 c) which would indicate differences of SSD patterns between mating modes. Cranial size and SSD-patterns do differ between mating modes (Tab.3 d). Pairwise comparisons showed that male and females of amplecting species differ (Z=2.04, p=0.019) in cranial size while this is not the case for circle dancers (Z=-1.76, p=0.96). For the humerus, the same pattern applies but the effect size between amplecting males and females is only close to significance (Z=1.53, p=0.066).