The longer the better: evidence that narwhal tusks are sexually selected

Abstract

Once thought to be the magical horn of a unicorn, narwhal tusks are one of the most charismatic structures in biology. Despite years of speculation, little is known about the tusk's function, because narwhals spend most of their lives hidden underneath the Arctic ice. Some hypotheses propose that the tusk has sexual functions as a weapon or as a signal. By contrast, other hypotheses propose that the tusk functions as an environmental sensor. Since assessing the tusks function in nature is difficult, we can use the morphological relationships of tusk size with body size to understand this mysterious trait. To do so, we collected morphology data on 245 adult male narwhals over the course of 35 years. Based on the disproportional growth and large variation in tusk length we found, we provide the best evidence to date that narwhal tusks are indeed sexually selected. By combining our results on tusk scaling with known material properties of the tusk, we suggest that the narwhal tusk is a sexually selected signal that is used during male–male contests.

1. Introduction

Protruding from the head and reaching nearly 3 m in length, the tusks of narwhals (Monodon monoceros) are one of the most charismatic structures in biology. Despite immense speculation, the function of the tusk remains unclear. Based on the tusk's anatomy, a recent study proposed that narwhal tusks sense chemical changes in their environment [1]. By contrast, reports of head scarring, broken tusks and tusks impaled in the sides of male narwhals suggest that males use their tusks as weapons during aggression [2–4]. Observations of ‘tusking’, where two narwhals cross and rub their tusks together, suggest that the tusk is used for communication during intra- or intersexual interactions [4] (figure 1). Unfortunately, narwhals spend most of their lives in inaccessible areas of the Arctic which makes these behaviours nearly impossible to study [3]. Additionally, almost all male narwhals are tusked, while females rarely develop a tusk [2,5]. Importantly, although narwhal tusks are generally sexually dimorphic, this sexual dimorphism alone does not provide sufficient evidence for sexual selection, because other selective pressures, such as differences in foraging behaviour, can lead to sexual dimorphism as well (i.e. ecological selection; [6–8]). Since we currently have no way of directly assessing the function of the tusk in nature, we can use the morphological relationships of tusk size with body size to better understand the function of this exaggerated trait [9,10].

Figure 1.

This narwhal (Monodon monoceros) behaviour, called tusking, has been recorded between two males with a female present, and with two or more males with no female present. Visual observations and photographs of this behaviour were used to recreate this image.

The scaling relationship of a trait with body size has historically been used to provide evidence regarding the type of selection acting on that trait [11]. When comparing individuals of the same age, sexually selected traits often exhibit disproportional growth, i.e. hyperallometry [12]. That is, for a given body size, sexually selected traits are often larger than expected in the largest individuals [11,13–19]. Using a dataset of 35 narwhals of different ages, Best [20] proposed that narwhal tusks are sexually selected solely based on tusk hyperallometry. However, we now know that non-sexual traits, such as lizard cranial horns, can also scale hyperallometrically without being sexually selected [9,17,21]. Furthermore, Best [20] combined juveniles and adults in one analysis, which limits the conclusions because reaching sexual maturity influences trait growth (i.e. ontogenetic allometry [22]). Therefore, a first step towards understanding the function of the narwhal tusk should be to examine how the tusk scales with the body size of adults only (i.e. static allometry).

A second step that is not addressed in previous studies is to compare the variation between the tusk and a trait that is selected for non-sexual functions [9,10]. Sexual selection typically amplifies individual differences, because individuals with more resources are expected to produce larger than average traits, while individuals with less resources produce smaller than average traits [23,24]. Conversely, if a trait that is vital for survival is smaller or larger than average, individuals are expected to perform worse and suffer fitness costs [25]. Thus, when we compare a putatively sexually selected trait to a non-sexually selected trait, we expect the trait that is under sexual selection to be more variable than the trait selected for non-sexual functions [10]. Since some of these methods have already been used to understand the function of morphological traits in extinct species [10,26–29] and were validated in extant species [10], we can apply these methods to understand enigmatic traits such as the narwhal tusk.

Here, we studied the morphological scaling and variance of 245 adult male narwhals, collected from 1983 to 2018. First, we examined the scaling relationship between body size and tusk length and compared it to the scaling relationship between body size and our reference trait, fluke width. If the tusk is sexually selected, we expect a higher variance and a steeper scaling in comparison to our reference trait. If the tusk, however, is selected for non-sexual functions, we expect equal variance and a similar scaling to the reference trait.

2. Material and methods

(a). Data collection

Morphology data (body length, fluke width and tusk length) and sex were collected from narwhals in Greenland from 1983 to 2018. Samples and data were collected through the Greenland Institute of Natural Resources (GINR) using a narwhal specific sampling scheme prepared by the GINR. The sex of the animal was determined either by the reproductive organs or by the presence (male) or absence (female) of a tusk. Male narwhals were considered adults when they reached 400 cm [30]. For our analyses, we only included adult males because we wanted to investigate the scaling relationships of individuals of the same age (i.e. static allometry [22])––adding sexually immature individuals would change the properties of the scaling relationship (i.e. ontogenetic allometry [22]).

(b). Statistical analysis

We predicted that if the male tusk is sexually selected, the tusk would have a steeper scaling value (greater values of slope, β) when compared to the fluke width scaling value. To test this hypothesis, we first ran three separate models for each trait (tusk length and fluke width): (i) a linear model, (ii) a quadratic model and (iii) a nonlinear exponential model to determine which model best described our data. For each trait, we then compared the fit of the three models using AIC [31]. The model with the lowest AIC value was used as the best fit model (using the threshold of ΔAIC > 2 to discard the other models), and we used the parameters estimated by that model on further analyses. If the models tied (ΔAIC < 2), we used the simplest model. In all models, we used the size of the trait (i.e. tusk or fluke) as our dependent variable and body length as our independent variable. After model selection, we compared the slopes of models with their 95% confidence intervals. We conducted all analyses using base R for linear and quadratic models, the nlme package for the exponential model [32] and bbmle package for the AIC tests [33].

We also predicted that if the narwhal tusk is sexually selected, we expect greater variation in tusk length compared to the variation in fluke width. To test this hypothesis, we used two analyses. First, we calculated the coefficients of variation (CV) for each trait and bootstrapped the CV to have a measure of reliability. We bootstrapped the CV using a standard non-parametric formula by sampling our data with replacement and calculating the CV. We performed 10 000 iterations and calculated the confidence interval of the CV using the boot package [34]. Second, we built two linear mixed regression models with different error structures and compared how they fit the data. Standard regression models use a unitary error structure (e.g. variance of the entire dataset) to calculate the best fit models. However, some traits might have higher variance than other traits, and in these cases, models tend to be ‘heteroskedastic’. Thus, we also built a second model that use two error structures: one error structure taken from tusk length data and another error structure taken from fluke width data [35]. Both models use trait size as the dependent variable, and body length, the type of trait (i.e. tusk or fluke) and their interactions as the independent variables. In our second model, we used individual as a random effect because tusk length and fluke width came from the same animal. Error structure models were compared using the same AIC thresholds from the allometry analysis.

3. Results

Tusk length was best fit by a quadratic model, whereas fluke width was best fit by a linear model (table 1 and figure 2a). Based on the best fit models, tusk length increased with body size with a strong, hyperallometric slope (β₁ = 9.06 [5.36; 12.75, 95% CI], β₂ (or x²) = −0.009 [−0.013; −0.005]; figure 2a). Interestingly, only three narwhals were larger than 5.2 m, and all these individuals had very small tusks (figure 2a). To assess the influence of these three individuals in our analysis, we reran the models without these three individuals, but results were similar (β₁ = 7.26 [1.25; 13.26, 95% CI], β₂ = −0.007 [−0.013; −0.0005]; figure 2a, dashed line). By contrast, fluke width increased with body size with a much shallower slope (β₁ = 0.23 [0.16; 0.29]; figure 2a). That is, larger males allocate relatively more into tusk length than smaller males. Our reference trait (fluke width), on the other hand, also increased as body size increased, although with a much shallower slope (table 1 and figure 2a). Furthermore, tusk length had a higher CV than fluke width (figure 2b), and the residuals of tusk length were also larger than the residuals of fluke width (table 1 and figure 2c).

Table 1.

Model comparisons for tusk length allometry, fluke width allometry and variance error structures. For each model, the number of parameters (k), the log-likelihood (ll), the Akaike information criterion (AIC) and the Akaike weight (w) are reported. Linear model: y ∼ a + b × x; quadratic model: y ∼ a + b × x + c × x²; exponential model: y ∼ a + exp (b × x).

	k	log-likelihood	AIC	ΔAIC	w
(a) tusk length allometry
quadratic model	4	−649.0	1306.1	0.0	1
linear model	3	−658.8	1323.6	17.6	<0.0001
exponential model	3	−664.4	1334.8	28.8	<0.0001
(b) fluke width allometry
linear model	3	−432.1	870.1	0.0	0.550
quadratic model	4	−431.5	870.9	0.8	0.363
exponential model	3	−433.9	873.8	3.7	0.088
(c) variance error structure
two error structures	8	−1095.4	2206.8	0.0	1
common error structure	7	−1156.2	2326.5	119.7	<0.0001

Figure 2.

(a) The scaling relationship between narwhal (Monodon monoceros) body size and tusk length (solid red) demonstrates the steep scaling of the tusk (y = −2036.5 + 9.06 × x − 0.009 × x²). When we removed three outliers that likely drive the downward slope of our data, an asymptote of tusk allocation appears (dashed red; y = −1640 + 7.26 × x − 0.007 × x²). Fluke width (grey), on the other hand, demonstrates the shallow scaling of a non-sexually selected trait (y = 3.689 + 0.23 × x). (b,c) Coefficients of variation and residual variation for tusk length (red) and fluke width (grey) demonstrates considerable variation in tusk length, but not fluke width, consistent with the hypothesis that the tusk is sexually selected.

4. Discussion

The hyperallometric scaling and large variation in the tusks supports the hypothesis that narwhal tusks are sexually selected (figure 2; electronic supplementary material, figure S1 and table S1). Sexually selected signals used in male–male competition are more likely to exhibit hyperallometry when compared to other sexually selected traits, because the information being signalled is simple: ‘I am bigger than you’ [9]. To convey this message, males exaggerate the size of their signals which facilitate the detection of size discrepancies between individuals, reducing the likelihood of engaging in potentially dangerous fights [9]. Using the tusk as a signal might explain the tusking behaviour observed in male narwhals (figure 1; [4]). Tusking may be a ritualized behaviour used to assess the fighting ability of opponents and avoid potentially costly fights [36]. In this scenario, fully fledged aggression should be rare, because competitors should attempt to settle contests through signalling [37,38].

However, using the tusk purely as a signal does not explain why head scarring and broken tusks occur. Adult male narwhals possess significantly more scars on their heads compared to juvenile males and females [3,4], similar to the scarring found in other cetaceans that are known to fight [39,40]. Similar injury patterns are also common in other animals that possess signals that also function as weapons [36,41,42]. Additionally, 40–60% of adult males have damaged or broken tusks [2,4], which should be common if narwhals used their tusks in fights [43]. Interestingly, tusks cannot withstand ‘ramming’ forces, which suggest that the tusk is not used for direct stabs [44,45]. However, tusks can withstand lateral strikes without breaking, which might explain the head scarring [44,45]. Considering that the tusk length scaling reaches an asymptote, then tusk length may be constrained by breaking when becoming too long [46] or there are large metabolic costs incurred when individuals produce and maintain massive tusks (but see [47]). Indeed, biomechanical limits to the size and elaboration of sexually selected traits are common [48–50]. Thus, when combing our results of tusk hyperallometry with the known material properties of narwhal tusks, it suggests that the tusk is likely used as an aggressive signal during contests.

An alternative hypothesis is that the tusk functions as a signal during mate choice [9]. Signals used in mate choice should also scale hyperallometrically based on a similar logic to aggressive signals. For example, male fiddler crab claws scale hyperallometrically and are used both in mate attraction and as aggressive signals during territorial disputes [51]. Moreover, if the tusk plays a role in mate choice, we may expect that the size of the tusk gives information about the quality of the wielder. If only the highest quality males produce and adorn larger-than-average tusks, then it can also serve as an honest signal to females or males (i.e. handicap signal, [23]). However, little is currently known about mating behaviour and the role of mate choice in narwhals.

Importantly, despite being sexually selected, the tusk might also possess other functions outside of sexual signalling or aggression. For example, the sexually selected claws of many crustaceans function as weapons during territorial contests, but they also can function as signals in male–male combat and female mate choice [41,52–54]. Additionally, the same claws play a role in chemical communication and prey capture [55,56]. In narwhals, the tusk might also possess other functions related to sensing the environment and in rare cases, catching prey [1]. Thus, although we have supported the hypothesis that the tusk is sexually selected, large tusks may benefit narwhals in non-sexual acts that we are currently unaware of. Proclaiming that a trait has a single function can hinder our understanding of trait evolution, because the multiple hypothesis for a trait’s function are not always mutually exclusive.

Overall, our evidence supports the hypothesis that the tusk functions both as a sexually selected weapon and sexually selected signal during male–male contests. However, further evaluations of the narwhal's ecology are warranted. Because female narwhals sometimes develop a tusk, investigations into anatomical, morphological and behavioural differences between tusked males and tusked females may shed light on the tusk's importance during non-sexual functions. Furthermore, population and seasonal differences into tusk investment may reveal the tusks relative importance during pre- and post-copulatory mating. Lastly, continuing efforts to collect behavioural data on narwhals (e.g. using drones) will provide new insights into the social structure and behaviour of this unique species. Although we do not discard other possibilities, based on the large amount of morphological data we gathered here, we suggest that the narwhal tusk likely functions as a signal and a weapon during aggressive disputes.

Data accessibility

R scripts will be made available upon request. Raw data have been archived in the Dryad Digital Repository: https://doi.org/10.5061/dryad.s1rn8pk43 [57].

Authors' contributions

This study was conceived by Z.A.G. and A.V.P.; E.G. and M.P.H.J. collected the data; Z.A.G. and A.V.P. analysed the data and drafted the manuscript. All authors edited the manuscript. All authors approved the final version of the manuscript and agree to be held accountable for the content therein.

Competing interests

We declare we have no competing interests.

Funding

We received no funding for this study.