Sexual selection and species recognition promote complex male courtship displays in ungulates
Data files
Apr 29, 2024 version files 7.39 KB
Abstract
Identifying the evolutionary drivers of sexual signal complexity is a key challenge in the study of animal communication. Among mammals, male bovids and cervids often perform elaborate gestural displays during courtship, consisting of ritualized movements of various parts of the body but the causes underlying interspecific variation in complexity of such displays remain poorly understood. Here we apply the comparative method to investigate which factors may have either promoted or constrained gestural repertoire size.
We found that sexual selection was a strong predictor of gestural display complexity in male bovids and cervids. Repertoire size was positively correlated with breeding group size, an indicator of the intensity of sexual selection on males. Moreover, repertoires were larger in species adopting non-territorial and lek breeding mating systems than in species adopting resource-defence territoriality, a finding that can be explained by more emphasis on direct benefits than indirect benefits in resource-defence systems, where male mating success may also be less skewed due to difficulty in monopolising mates.
The results also indicate that gestural repertoire size was positively correlated with the number of closely-related species occurring in sympatry. This is consistent with display complexity being selected to facilitate species recognition during courtship and thereby avoid interspecific hybridization. At the same time, repertoire size was negatively associated with male body mass, possibly due to the energetic and mechanical constraints imposed on movements in very large species. By contrast, we found no evidence that the habitat drives selection for complex gestural courtship displays.
README: SEXUAL SELECTION AND SPECIES RECOGNITION PROMOTE COMPLEX MALE COURTSHIP DISPLAYS IN UNGULATES
https://doi.org/10.5061/dryad.msbcc2g5s
Description of the data and file structure
Identifying the evolutionary drivers of sexual signal complexity is a key challenge in the study of animal communication. Among mammals, male bovids and cervids often perform elaborate gestural displays during courtship, consisting of ritualized movements of various parts of the body but the causes underlying interspecific variation in complexity of such displays remain poorly understood. Here we apply the comparative method to investigate which factors may have either promoted or constrained gestural repertoire size.
We found that sexual selection was a strong predictor of gestural display complexity in male bovids and cervids. Repertoire size was positively correlated with breeding group size, an indicator of the intensity of sexual selection on males. Moreover, repertoires were larger in species adopting non-territorial and lek breeding mating systems than in species adopting resource-defence territoriality, a finding that can be explained by more emphasis on direct benefits than indirect benefits in resource-defence systems, where male mating success may also be less skewed due to difficulty in monopolising mates.
The results also indicate that gestural repertoire size was positively correlated with the number of closely-related species occurring in sympatry. This is consistent with display complexity being selected to facilitate species recognition during courtship and thereby avoid interspecific hybridization. At the same time, repertoire size was negatively associated with male body mass, possibly due to the energetic and mechanical constraints imposed on movements in very large species. By contrast, we found no evidence that the habitat drives selection for complex gestural courtship displays.
Dataset_Sexual_Selection_Species_Recognition_Gestural_Courtship_Ungulates_-_Sheet1
Data include one single entry for each species of Bovidae and Cervidae considered for the study. Data include: (i) gestural display complexity, intended as the overall species-specific repertoire of gestural courtship displays (described in the methods); (ii) average Breeding Group Size, as calculated from literature; (iii) average species-specific Habitat Openness, calculated as the average proportion of openness across habitat types listed by the IUCN Red List for each species (openness proportion defined by Stankowich et al. 2016); (iv) average Male Body Mass, derived from literature, in Kg; (v) Research Effort, intended as the total number of studies on behaviour for each species, derived from ISI Web of Knowledge; (vi) Degree of Sympatry, defined as the total number of species in the same tribe showing any overlap in distribution range, based on distribution maps from the IUCN Red List; (vii) Male Mating Strategy, defined as Lek, Territorial, or Non-Territorial, based on literature reported in the Methods.
Supplementary Information
The Word doc stored in Zenodo contains Appendix I, II, and data sources.
Sharing/Access information
n/a
Code/Software
n/a
Methods
Scoring gestural display complexity as the response variable
Data on gestural courtship displays of bovids and cervids were collected from peer-reviewed publications and scientific books in English, French, German, Italian, and Spanish (data sources listed in Appendix II). We only considered descriptions that detailed full courtship behavior sequences leading to copulation and excluded sources which only mentioned specific displays without describing an entire courtship sequence. This resulted in a dataset of 73 out of 136 bovid species and 21 out of 47 cervid species. Although we failed to find descriptions of courtship behavior for some species, the dataset is representative of the overall diversity in that it includes at least one species from each genus, except for the poorly studied monospecific genus Pseudoryx (Bininda-Emonds et al. 2007). Domesticated species were excluded from the study because human intervention is likely to have modified their behavioral traits.
We used the repertoire size of gestural courtship displays as a quantitative measure of species-specific display complexity (Ord & Blumstein 2002; Dunn & Smaers 2018; Miles & Fuxjager 2018b). Repertoire size was scored as the total number of distinctive body movements performed by males during courtship, henceforth referred to as “display components” (Table 2; Appendix I). Definitions and classification of display components were based on the seminal work of Fritz R. Walther (1974, 1984). Some display components (e.g. ear and horn orientation) are subtle modifiers of other movements (such as head-up and low-stretch postures; Walther 1984; Appendix I) and are not consistently described across species; these were excluded from the calculations. Also, following Walther (1984), executive behaviors with particular functions were not scored as display components; these include licking, smelling, grooming, and naso-genital contact with Flehmen response (olfactory investigation of sexual receptivity). A small number of gestural courtship display components involved movements aimed at enhancing the expression of another behaviour, e.g. squatting during urine spraying (Walther 1984; Schaller 2000). We conducted the statistical analyses also without adding them to the final gestural repertoire score, and obtained qualitatively similar results (not shown).
Independent variables
The explanatory variables included in the study included mating system, which was scored as a categorical variable, and the following continuous variables: breeding group size, degree of sympatry with closely related species, habitat openness, average male body mass (in kg), and research effort (data sources listed in Appendix II). Mating system was classified as either resource defence territoriality, lek territoriality, or non-territorial female defence.
Males in a few ungulate species are reported to adopt alternative mating systems in different populations, and sometimes also within the same population (Bro-Jorgensen 2011; Corlatti & Lovari 2023). We assigned these species to the predominant mating system in the population for which the repertoire size had been calculated. For example, male red deer (Cervus elaphus) defend resource-based territories in some areas (Carranza, Alvarez & Redondo 1990), but descriptions of courtship displays were only available for populations where non-territorial female defence was the main mating system, and this species was therefore classified as non-territorial. In species that lek, lek, and resource-based territoriality usually co-exist in lekking populations; however, mating success is usually skewed in favour of lekking males (Isvaran & Ponkshe 2013; Isvaran 2021) and this affects also the mating success of resource defenders (Bro-Jørgensen & Durant 2003). Even in populations that are not forming classical leks, more successful territories are often clustered and can be considered as lek precursors or “exploded leks” (Bro-Jorgensen 2008). Therefore, species that form leks in at least some populations were classified as lek territorial.
The degree of sympatry with closely related species was quantified as the number of other species from the same tribe with whom distribution ranges overlapped. Bovid and cervid tribes were defined following Vrba & Schaller (2000) and all corresponded to monophyletic groups in the phylogeny used for the comparative analysis (Bininda-Emonds et al. 2007). We chose to focus on sympatry on the tribal rather than the generic level because interspecific hybridization has been observed also between bovid species not in the same genus (e.g. Jorge, Butler & Benirschke 1976; Douglas et al 2011). Overlap in geographic distribution was assessed visually by plotting distribution polygons for all species within the tribe using QGIS 3.4.3 (QGIS Development Team 2019). Any observable overlap in geographic range was considered as evidence for sympatry (Santana et al. 2013). Polygons of distribution range were obtained from the IUCN Red List database (www.iucnredlist.org). We scored the degree of sympatry based on current distribution ranges because spatial data on historical occurrence is inadequate for the majority of the species included in this study.
Following Stankowich & Campbell (2016), habitat openness was scored as the probability of detection for terrestrial mammals in each of the seven main habitat categories in the IUCN Red List classification scheme (www.iucnredlist.org),: (i) 0.10 tropical rainforest; (ii) 0.20 temperate forest; (iii) 0.30 wetland; (iv) 0.50 shrubland; (v) 0.70 grassland (tropical and temperate); (vi) 0.80 rocky areas; and (vii) 0.95 deserts. Scores were assigned only to habitats reported as “suitable”, and the overall species-specific habitat openness score was calculated as the average detection probability across all habitat categories.
Because gestural repertoire sizes calculated from literature may conceivably be biased towards more well-studied species, we controlled for research effort using the number of publications mentioning the Linnean binomial name of each species in the ISI Web of Knowledge (www.webofknowledge.com) between 1960 and 2018 (as no accounts of courtship behavior have been published since then). The search was restricted to the categories likely to include behavioral accounts, i.e. (i) Zoology; (ii) Behavioral Sciences; (iii) Ecology; and (iv) Evolutionary Biology.
Statistical analyses
All analyses were conducted in R v. 3.5.2 (R Development Core Team 2019) with the packages ape and caper loaded in the workspace (Orme et al. 2018; Paradis et al. 2019). We used phylogenetic least squares (PGLS) regressions to identify statistically significant predictors of gestural repertoire size. The PGLS methods accounts for autocorrelations in the dataset generated by shared ancestry by including phylogeny as a variance-covariance matrix in the error structure of a least squares regression models (Harvey & Pagel 1991; Housworth, Martins & Lynch 2004). The phylogeny for this study was derived from the ultrametric molecular tree of mammals in Bininda-Emonds et al (2007) and pruned to include only the species included in the dataset. This phylogeny was selected as it offers the best species coverage for both cervids and bovids, incorporating both molecular and morphological data. Branch lengths were scaled with Pagel’s lambda set to maximum likelihood (Freckleton, Harvey & Pagel 2002) as this transformation best fitted the dataset after graphical comparisons with delta and kappa estimators (using the profile.pgls function in caper; Orme et al. 2018).
Gestural repertoire size was entered as the response variable in PGLS models together with the following explanatory variables: breeding group size, degree of sympatry, habitat openness, male body mass, research effort (all continuous) and mating system (categorical). Male body mass and breeding group size were log-transformed (using the natural logarithm) prior to analyses in order to meet the assumptions of residual normality and homoscedasticity (graphically checked using the plot.pgls function in the package caper; Orme et al. 2018). Model simplification was conducted by progressive removal of non-significant predictors in order of least significance (p ≤ 0.05; Murthaugh 2014). The results presented here pertain to the final model including only significant predictors; statistics for non-significant predictors were obtained by separately adding each of them separately to the final model. Variance inflation factors (VIFs) were calculated to estimate multicollinearity between independent variables. All VIFs were ≤ 2.04, and thus well below the commonly accepted threshold of concern (5-10; McClave & Sincich 2003). We moreover tested for correlations between gestural repertoire size and each of the predictors separately in bivariate models.
Finally, to further explore the Sexual Selection hypotheses, we also tested for sexual size dimorphism (SSD) (measured as male body mass:female body mass) as a predictor of courtship display repertoire size. SSD is commonly used as an indicator of sexual selection on males (e.g. Cassini 2020), however correlation between SSD and both breeding group size and male mating strategy prevented us from including all three variables in the same analysis due to multicollinearity issues.