Intraspecific male combat behaviour predicts morphology of cervical vertebrae in ruminant mammals

Abstract

Cranial weapons of all shapes and sizes are common throughout the animal kingdom and are frequently accompanied by the evolution of additional traits that enhance the use of those weapons. Bovids (cattle, sheep, goats, antelope) and cervids (deer) within the mammal clade Ruminantia are particularly well known for their distinct and varied cranial appendages in the form of horns and antlers, which are used as weapons in intraspecific combat between males for access to mates. Combat in these species takes many forms, including head-on collisions (ramming); stabbing an opponent's head or body with horn tips (stabbing); rearing and clashing downwards with horns (fencing); or interlocking antlers or horns while vigorously pushing and twisting (wrestling). Some aspects of weapon and skull morphology have been linked to combat behaviours in bovid and cervid species, but the contribution of postcranial structures that support these weapons, such as the neck, has not been explored. To investigate the role of the neck in intraspecific combat, we quantified biomechanically relevant linear variables of the cervical vertebrae (C1–C7) from males and females of 55 ruminant species. We then used phylogenetic generalized least-squares regression to assess differences among species that display primarily ramming, stabbing, fencing and wrestling combat styles. In males, we found that wrestlers have longer vertebral centra and longer neural spines than rammers, stabbers or fencers, while rammers have shorter and wider centra and taller neural spine lever arms. These results suggest a supportive role for the cervical vertebrae in resisting forces generated by male–male combat in ruminant mammals and indicate that evolutionary forces influencing cranial weapons also play a role in shaping the supporting anatomical structures.

1. Introduction

Animal clades from insects and crustaceans to dinosaurs and mammals have evolved weapons that are used in contests to directly or indirectly increase reproductive success [1]. The presence of these sexually selected weapons is often associated with parallel changes in other traits, some of which may compensate for any performance costs imposed by the development or use of the weapons [2,3]. Some compensatory traits involve functional or morphological trade-offs (i.e. reduced antenna length to compensate for increased weapon size in horned beetles [4]), while others appear to directly support the successful use of weapons (i.e. increased forefemur and head size in the same beetle species enhance the performance benefit of the weapon [5]). While traits that compensate for or support the use of sexually selected weapons have been studied in insects [4,5] and crustaceans [6], the role of supporting structures is less well understood in vertebrates broadly and mammals in particular.

Perhaps the most striking examples of sexually selected weapons in mammals belong to ruminants, the clade that includes the families Bovidae (cattle, sheep, goats, antelope) and Cervidae (true deer) [7]. Having undergone two probable adaptive radiations [8], the roughly 130 extant bovid species and 50 extant cervid species now display an astonishing diversity of body sizes, ecological niches, social behaviours and morphologies [9]. Ruminants are noted for the extreme variation in cranial appendages across the clade, including horns in bovids and antlers in cervids [10], which are used as weapons in intraspecific competition between males for access to mates [11–14]. Bovids and cervids also exhibit varied social behaviours, ranging from monogamous mate-pairs to groups of hundreds or even thousands of individuals with strict social dominance hierarchies, intensely contested mating territories, ritualized displays of aggression and often physical combat between competing males [15,16]. These contests can result in mortality or injury of the combatants [13,14,17] and increased susceptibility to mortality through predation and malnutrition in species in which males defend territories, lek or guard females [18–20]. Despite the steep costs of intraspecific competition, the rewards for dominant males are considerable, as male reproductive success in these species depends heavily on the ability to win fights [14].

Intraspecific combat in male bovids and cervids takes many forms, including head-on ramming (e.g. bighorn sheep) [12], stabbing (e.g. impala, mountain goats) [11,21], fencing and clashing with horns (e.g. ibex, oryx) [22,23], and interlocking antlers or horns for vigorous wrestling (e.g. deer, elk and many antelopes) [11,14] (figure 1). The dramatic nature of these contests and the variety of behaviours exhibited across the clade has led many researchers to search for associated traits that allow ruminants to cope with the mechanical demands of combat. Horn and antler shape in male bovids and cervids is correlated with broad trends in social behaviour [25] and with the use of specific behaviours when fighting [26,27]. The material properties of the weapons may also play a role in resisting forces generated by combat, including those of the keratin sheath [28] (but see [29]) and bony horn core [30,31] of bovid horns, and the antler of red deer [32].

Figure 1.

Phylogeny of ruminant species included in this study, adapted from [24], with major clades identified in brackets. Fighting categories are displayed at the tips. (Online version in colour.)

Whereas the elaborate cranial appendages have drawn the attention of both female ruminants and most researchers, very little is known about the role of any body part distal to the skull in generating and resisting fighting forces. Postcranial traits that support the use of sexually selected weapons have been identified in some insects [4,5] but are largely unexplored in ruminant mammals. Previous authors have suggested that substantial fighting energy must be absorbed by the body musculature [29] and that the torque generated by the use of the horns must be opposed by the neck muscles [23,33], but these hypotheses have not, to our knowledge, been quantitatively tested.

To determine whether the ruminant neck displays traits that support the use of the horns and antlers in intrasexual combat, we explore the relationships between cervical vertebra morphology and fighting behaviour across bovids and cervids. Cervical vertebrae provide attachment surfaces for muscles and ligaments that support the head and flex, extend and rotate the head and neck [34]. If the ruminant neck has evolved adaptations to generate and resist mechanical forces during intraspecific combat, we predict that cervical vertebra morphology will be correlated with the use of specific fighting behaviours across bovid and cervid species.

To test the relationship between fighting behaviour and cervical vertebra morphology in bovids and cervids, we used biomechanically relevant linear measurements of all seven cervical vertebrae from adult male and female specimens across the clade. Females have antlers in only one species of cervid, and the presence of horns in female bovids is highly variable. When present, female horns are often reduced in size and less complex in shape than male horns. While male weapons are influenced by sexual selection, the primary function of horns in female bovids appears to be anti-predator defence [35]. Given the different role for female weapons, and the lack of ritualized combat behaviour among females, we expect that aspects of vertebral morphology will be more strongly associated with fighting behaviour in males than in females.

2. Methods

(a). Sampling and measurements

We measured vertebrae C1–C7 of 130 ruminant specimens from 55 species in collections at the Museum of Comparative Zoology (MCZ), Cambridge, MA, USA; the American Museum of Natural History (AMNH), New York, NY, USA; the United States National Museum of Natural History (USNM), Washington, DC, USA and the Field Museum of Natural History (FMNH), Chicago, IL, USA. We used only adult specimens with complete cervical vertebral columns and identified sex information (see the electronic supplementary material, table S1)—in many species, only one male and one female were available, although in a few species, we obtained two or three male and female specimens and computed the species mean for each variable.

We aimed to capture biomechanically relevant aspects of cervical vertebral morphology using linear measurements (figure 2), which we collected to the nearest tenth millimetre using Mitutoyo digital calipers. We measured the length, width and height of the vertebral centrum (or vertebral arches on C1), the width of the prezygapophyses, the craniocaudal length and lever arm distance of the neural spine, and the craniocaudal length and lever arm distance of the transverse processes (figure 1). Not all features were present on all vertebrae, and the atlas (C1) and axis (C2) in particular are quite different from the remaining cervical vertebrae, reflecting their specialized roles in flexing and extending (C1) and rotating (C2) the skull [34].

Figure 2.

Linear measurements of cervical vertebra features. Measurements were taken based on feature availability on all seven cervical vertebrae; examples shown here on C1, C2, C6 and C7. LVA, length of ventral arch; WVA, width of ventral arch; preZD, prezygapophyseal distance; CL, centrum length; CW, centrum width; CH, centrum height; TPL, transverse process length; TPLA, transverse process lever arm; vTPLA, ventral transverse process lever arm; NSL, neural spine length; NSLA, neural spine lever arm.

Most biomechanical hypotheses pertaining to the spine have been tested in thoracic or lumbar vertebrae and our predictions for cervical vertebra shape are guided by these studies. Longer centra allow for increased flexibility and lateral bending, while shorter centra are more rigid [36–38]. Wider centra reduce lateral flexibility and increase compressive strength, while taller centra limit dorsoventral flexibility [36,39–41]. Craniocaudally longer neural spines provide more attachment surface for the nuchal ligament and the dorsal musculature [42], and taller neural spines reduce dorsoventral flexibility and provide greater leverage for extensor muscles [36,41]. Likewise, craniocaudally longer transverse processes increase the attachment area available for lateral and ventral muscles, while longer transverse process lever arms restrict lateral flexibility and increase leverage for muscles used in lateral bending and flexion [36,37,41].

(b). Body mass and fighting behaviour

We used sex-specific species mean body mass from the published literature to account for body size differences in our sample (electronic supplementary material, table S1). Species were placed into one of four fighting style categories based on literature accounts of the primary fighting behaviours displayed between conspecific males (figure 1). The ‘ramming’ category includes species that make brief, frontal contact with the base of the horns or skull before disengaging (e.g. the bighorn sheep Ovis canadensis). The ‘wrestle’ category includes species that lock horns or antlers and twist or push against their opponent for a prolonged period of engagement (e.g. the white-tailed deer Odocoileus virginianus or the impala Aepyceros melampus). The ‘fence’ category includes species that make downward or lateral clashing movements with horns and do not lock horns or do so only briefly before disengaging for another clash (e.g. the scimitar-horned oryx Oryx dammah). The ‘stab’ category includes species that attempt to slash or stab with the horn tips towards an opponent's head or flanks but do not lock horns or make skull contact (e.g. the dik-dik Madoqua kirkii and the American mountain goat Oreamnos americanus).

(c). Phylogenetic comparative methods and statistical analyses

To account for the evolutionary non-independence of body size, vertebral morphology and behaviour, we used phylogenetic generalized least-squares (PGLS) analyses to determine the relationship between each linear variable, with body mass and fighting style as covariates. We used a time-calibrated tree of mammals [24] pruned to the 55 ruminant species for which we had measurements from at least one male and one female specimen (figure 1).

Prior to analyses, we computed mean measurements for species in which we had data from multiple specimens and log-transformed all linear measurements and body mass values. We then performed a PGLS analysis for each linear measurement via the R package caper, using the maximum-likelihood estimation of Pagel's λ to transform branch lengths [43]. To avoid bias owing to collinearity of body size with other variables, we opted to include body mass as a separate term in each PGLS model rather than using residuals from a regression of the vertebral measurement on the body mass [44]. Our models therefore used the formula: vertebral measure ∼ body mass + fighting category. This allowed us to account for the effect of phylogeny on body mass, the effect of phylogeny on vertebra shape and the effect of body mass on vertebra shape within a single model.

We conducted separate PGLS analyses of vertebral shape and fighting style for males and for females. We also conducted a third set of PGLS analyses to test whether vertebral morphology was associated with the presence of horns in females, as female weapons are not present in all ruminant species. We used female horn presence/absence data from the published literature [25] and included it as a categorical variable in conjunction with the body mass.

In the PGLS method implemented in the caper package, one level of our categorical variable (fighting style) is treated as a reference category. Data points in this category are used to estimate the model intercept (the parameter β₀ in the regression equation ŷ = β₀ + β₁ × x), and the model then calculates the residual error explained by each of the other categories in relation to the reference category. The function then generates parameter estimates associated with each category, as well as the standard error and p-value of each estimate. This is not the same as conducting between-group post hoc significance tests for each group, which would substantially increase the number of statistical tests. Instead, this framework allowed us to determine which vertebra measurements differed between the reference category and any other behavioural category. Based on prior assumptions about the group that was likely to exhibit the greatest differences [45], we designated ‘ramming’ as the reference category for all analyses. We then identified the lowest p-value associated with any fighting category in the analysis of each variable and evaluated that p-value as an indication of the strength of association between the vertebral feature and any kind of difference in fighting behaviour.

To avoid spurious associations owing to the high number of variables and individual analyses (44 for each sex), we used a standard Bonferroni correction to set the α criterion level to 0.0011 (0.05/44 = 0.0011). The standard Bonferroni correction is considered fairly conservative and results in a reduced probability of type 1 error (false positive) at the expense of type 2 error (false negative) [46,47]. In our effort to find meaningful associations between vertebral morphology and fighting style, we opted to use this conservative α criterion as a way to identify models with the strongest likelihood of an association between variables.

In accordance with the recommendation of the editors of The American Statistician, we have discontinued the practice of referring to results as ‘statistically significant’ or ‘statistically insignificant’ based on the relation of a p-value to a defined cut-off [48]. They and other authors in a recent issue of the journal strongly argue that p-values must be reported as continuous values and evaluated within the context of the study and its limitations [48,49]. We found this to be an effective framework for reporting the results of this study, considering our goal of identifying which of the many cervical vertebral traits measured would best predict fighting style and would be good candidates for further, more detailed analysis. We therefore considered that analyses for which the lowest p-value obtained from the model estimate of one of the fighting categories was less than 0.0011 suggest a plausible association between that variable and fighting style in ruminants.

3. Results

Body mass was strongly associated with vertebral morphology in PGLS analyses of all variables for both males and females (p < 0.00000001 in all cases). By using the conservative α criterion discussed above, we identified a handful of variables that were tightly linked with differences in fighting style between groups in addition to body mass (table 1 and figure 3). Centrum length was associated with fighting style in the mid and lower cervical vertebrae in both males and females (males: C3, C5, C6; females: C2–C6). Centrum width was associated with fighting style only in C2 in males. In C1 (which lacks a centrum), the width of the ventral arch was associated with fighting style in both males and females (C1 vertebral arch length and width are reported as the first square on lines G and I of figure 3, respectively). Centrum height does not appear to be strongly associated with fighting style in any vertebrae in males or females. Prezygapophyseal distance of C1 and C2 (essentially a measure of the width of the articular surface) did not meet our p-value threshold in either males or females. The craniocaudal length of the C2 neural spine was associated with fighting style in males; neural spine length was not associated with fighting style in any other vertebrae. Likewise, only one measure of lever arm height was associated with fighting style across all of the cervical vertebrae—the C7 neural spine lever arm in males. No characteristics of the transverse processes, including the craniocaudal length or lever arm distance, were associated with differences in fighting style in any of our models. The maximum-likelihood estimate of the phylogenetic signal in the model residuals was 0 in all analyses of variables that fell below our p-value threshold, with the exception of C2 centrum width in males (λ = 0.6).

Table 1.

Results of PGLS analyses of all variables where the lowest p-value associated with any fighting category was less than 0.0011 (see figure 3). (p(body mass) = p-value associated with body mass estimate. Body mass estimate, the parameter adjusting the predicted variable value based on body mass. Minimum p(fight style), the lowest p-value associated with any fighting style category. R²_, overall variance explained by the model. ML λ, maximum-likelihood estimate of Pagel's λ, used to adjust the amount of phylogenetic signal in the model residuals. Ram estimate (intercept), the parameter associated with a data point belonging to the ramming group, designated as the model intercept. Wrestle estimate, stab estimate and fence estimate, the parameters adjusting the predicted variable value relative to the intercept for data points in each fighting category, respectively. For example, in the model predicting C2 centrum width, the wrestling, stabbing and fencing parameter estimates are all negative, meaning they are lower than the ramming parameter estimate (intercept), and indicating that rammers have higher values for the centrum width measurement (i.e. wider centra) than all other categories.)

vertebra	variable name	p(body mass)	body mass estimate	minimum p(fight style)	ram estimate (intercept)	wrestle estimate	stab estimate	fence estimate	R2	ML λ
(a) male PGLS analyses
C1	ventral arch width	<1 × 10−10	0.298	0.0000	1.130	−0.095	−0.128	−0.112	0.93	0.00
C2	centrum width	<1 × 10−10	0.345	0.0002	0.849	−0.057	−0.118	−0.084	0.91	0.60
C2	neural spine length	<1 × 10−10	0.220	0.0009	1.320	+0.110	−0.020	+0.078	0.81	0.00
C3	centrum length	<1 × 10−10	0.179	0.0002	1.260	+0.132	−0.025	+0.054	0.76	0.00
C5	centrum length	<1 × 10−10	0.199	0.0007	1.180	+0.115	−0.035	+0.036	0.79	0.00
C6	centrum length	<1 × 10−10	0.222	0.0003	1.070	+0.114	−0.025	+0.031	0.84	0.00
C7	neural spine lever arm	<1 × 10−10	0.386	0.0010	1.260	−0.115	−0.060	−0.087	0.87	0.00
(b) female PGLS analyses
C1	ventral arch width	<1 × 10−10	0.305	0.0004	1.100	−0.093	−0.094	−0.070	0.93	0.00
C2	centrum length	2.0 × 10−10	0.208	0.0006	1.300	+0.123	+0.011	+0.049	0.85	0.42
C3	centrum length	1.6 × 10−5	0.153	0.0006	1.280	+0.151	+0.001	+0.050	0.76	0.00
C4	centrum length	3.1 × 10−6	0.164	0.0003	1.230	+0.156	+0.006	+0.064	0.75	0.46
C5	centrum length	7.4 × 10−8	0.189	0.0002	1.160	+0.153	+0.024	+0.052	0.79	0.00
C6	centrum length	1.1 × 10−8	0.195	0.0001	1.070	+0.154	+0.029	+0.056	0.84	0.00

Figure 3.

Greyscale heatmap showing the range of the lowest p-value associated with a fighting category in each of the PGLS analyses performed using 44 variables (y-axis) across all seven cervical vertebrae (x-axis) in both (a) males and (b) females. The lightest grey squares correspond to the p-values of 0.05 and greater; progressively darker squares correspond to p-value ranges below 0.05; the black squares correspond to variables with p-values below our conservative Bonferroni cut-off of p = 0.0011.

In the entire set of female weapon presence/absence analyses, no cervical vertebra measurement was strongly associated with the presence or absence of female weapons (p > 0.01 for all analyses, well above our criterion for strong support) (electronic supplementary material, table S2).

Parameter estimates for each fighting category generated from the PGLS models for these variables indicate that in males, wrestlers have longer neural spines (C2) and longer centra (C3, C5–C6) relative to their body size than fencers, stabbers and rammers. Conversely, rammers have wider ventral arches (C1) and centra (C2) relative to body size than other groups. The neural spine lever arm of C7 is also taller in rammers than other groups and is smallest in wrestlers (table 1).

4. Discussion

We found that key aspects of cervical vertebra morphology are associated with differences in combat behaviour in ruminant taxa across the bovid and cervid radiations. The wide range of intraspecific competitive behaviour and a corresponding array of conspicuous cranial appendages in these animals has spurred many researchers to address the relationship between weapon morphology and fighting behaviour [11,12,25,26]. However, hypotheses about the role of the neck in resisting combat forces were limited to isolated species and had not been quantitatively tested [23,29,33,45]. Using a rigorous phylogenetic framework, we identified a potential set of postcranial adaptations related to intraspecific combat and cranial weapon use in bovids and cervids (table 1).

While the idea that animal weapons can spur morphological or behavioural modifications is not novel, the role of supporting structures in the evolution and use of these weapons is relatively unknown in mammals. In many insects with sexually selected weapons, additional correlated traits have been found to support the use of the weapons or compensate for the costs of using them [2,3,5]. Fighting success in flour beetles and bean bugs is significantly increased in males with an enhanced suite of traits—for example, male flour beetles require larger forefemurs, heads and prothraxes to bite and lift up opponents with their enlarged mandibles (weapons) [5]. Although there are no data evaluating the fighting success or the fitness costs of using sexually selected weapons across the 55 ruminant species in our morphological dataset, our analyses are a first step towards identifying a set of supportive traits that bolster the use of these weapons. Our results suggest that the morphology of the neck may have evolved to facilitate the disparate combat behaviours seen among bovids and cervids and illustrate a more cohesive anatomical picture of cranial weapon use and fighting behaviour in different species.

Species that engage in ramming behaviour typically experience repeated high-speed, short-duration impacts involving head-on contact at the base of the horns or skull [12]. In bighorn sheep, these impacts are estimated to produce up to 3000 N of force [29,33]. In our models, ramming species have shorter and wider vertebral arches (C1) and centra (C2) relative to body size than wrestlers, stabbers and fencers (table 1a). These results are consistent with the prediction that shorter and wider vertebrae in rammers may help absorb high compressive forces during impact while also resisting flexion, extension or lateral wrenching that could lead to injury in an off-centre collision [36,37,39,41]. Males in ramming species also have the tallest C7 neural spine lever arms (table 1a), which could reflect a need for greater neck extensor muscle leverage or an increased amount of dorsal muscle mass to effectively oppose accidental flexion during impact [33,45].

In contrast with the short, high-impact forces of ramming, wrestling species typically engage in longer bouts and use sustained horn contact to twist, push and wrench the opponent's neck side to side [11,25]. For example, male bushbuck antelope interlock their spiral horns and vigorously twist and push against each other for sustained periods [50], while white-tailed deer have convergently evolved a similar fighting style using interlocked antler tines [51]. We found that males in wrestling species have substantially longer and narrower vertebral arches (C1) and vertebral centra (C3, C5, C6) relative to body size than other fighting styles (table 1a). Longer and narrower centra in wrestlers probably allow for the intervertebral flexibility required by the extended engagement times and wide ranges of motion used by wrestling animals [22,52,53]. Wrestlers also have on average the shortest C7 neural spine lever arms relative to body size than rammers, stabbers and fencers, which probably accommodate less dorsal muscle mass and contribute to neck flexibility.

Fencing species typically employ repeated short lateral or downward clashes with the horns to the opponent's horns or skulls [22]. For example, male ibex rear up on their hind limbs and bring their horns down in an arcing motion, generating substantial downward torque on the neck [45] and probably requiring increased ventral muscle mass in order to resist neck extension [33]. Fencers have narrower C1 ventral arches and C2 centra than rammers, which is consistent with expectations, and narrower arches and centra than wrestlers, which is not (table 1a). However, vertebral centra in fencers are longer than in rammers, but shorter than in wrestlers—consistent with a need for intermediate stability and flexibility. Fencers also have intermediate C7 neural spine lever arm height, shorter than rammers but taller than wrestlers, and probably accommodating an intermediate amount of dorsal muscle mass. Because no analyses examining transverse process dimensions were well supported, our expectations that fencers would have increased area for ventral neck muscle attachment were not borne out.

Stabbing species typically attempt to stab or slash an opponent's head, limbs or flanks with the tips of the horns [25]. These fights involve less overall force on the neck, because the threat of injury comes from the sharp horn tips rather than blunt force or torque delivered with the horns or skull [50]. Some stabbing species such as duikers, rarely ever make horn contact during fights, preferring to fight with an ‘air cushion’ between them as they dodge in and out while brandishing the horns [50]. We found that males in stabbing species actually have the narrowest vertebral arches (C1) and centra (C2) of all fighting categories, as well as the shortest centra (C3, C5, C6) (table 1a). The generally narrow and short shape of the vertebral centra in stabbers may reflect the lack of forces acting on the neck compared with the other fighting styles [11].

Interestingly, we found no strong indication of a relationship between the transverse process morphology and the fighting style. The transverse processes serve as attachment points for intravertebral ligaments and muscles and vary considerably in relative length and lever arm distance from the centrum. In the thoracic and lumbar vertebrae, these processes provide leverage for horizontal movements and rotation, and differ based on locomotor behaviour in pinnipeds [41] and felids [54]. It may be that the cervical transverse processes in ruminants are influenced by factors other than those associated with combat.

Our analyses of vertebral shape in males and females differ in a few important ways. While centrum length and width of some vertebrae (or vertebral arch dimensions in C1) differ between fighting categories in both males and females, neural spine characteristics appear to differ notably among groups only in males (figure 3). We also found no strong relationship between any aspect of cervical vertebra morphology and the presence or absence of weapons in females, regardless of male fighting behaviour (electronic supplementary material, table S2). Although females possess cranial weapons in some species of bovid and one species of cervid, female weapons are typically smaller and less complex than male weapons, and their primary use is in anti-predator defence rather than the intraspecific competition [35,55]. Differences in male fighting style are associated with some aspects of female cervical vertebra shape—mostly centrum length—but the presence or absence of female horns is not. It appears that the functional signal in female neck vertebrae is not as strong as in males, which is not surprising given that female ruminants do not engage in the prolonged or ritualized combat behaviours shaped by sexual selection in their male counterparts [56,57] and therefore probably experience less extreme forces in the horns, skull and neck. Sexual dimorphism in body size and weapon size and shape is linked to mating strategy and reproductive behaviour throughout the ruminant clade [58–60]. We can now suggest that dimorphism related to intraspecific competition is also present in the cervical spine.

Overall, our study demonstrates a relationship between cervical vertebra morphology and fighting style in ruminants and identifies a set of postcranial traits that appear to support the use of sexually selected weapons in males. These vertebral features are just a small part of a complex suite of morphologies and behaviours that are related to the diverse ecologies, social structures and reproductive strategies that have evolved, sometimes convergently, among members of the ruminant radiation [8,9,11]. Although we purposely used strict criteria for selecting variables of potential biomechanical importance, it is clear that there are rich and varied relationships among vertebra characteristics, vertebra position, body size and behaviour that warrant further analysis. Considering the role of postcranial morphology in the evolution and maintenance of this eye-catching cranial weapon display will further illuminate the dynamic interplay of sexual selection, morphology and behaviour that influences intraspecific combat in ruminants.

Data accessibility

Morphometrics data and R analysis code are included in the electronic supplementary material.

Authors' contributions

A.V.L. collected the data, performed the analyses, prepared the figures and wrote the manuscript. E.R.D. advised on study design and appropriate data analysis and helped write and revise the manuscript.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by NSF GRFP award no. 1451512 to A.V.L.