New Geometric Morphometric Insights in Digital Taphonomy: Analyses into the Sexual Dimorphism of Felids through Their Tooth Pits

Featured Application

The identification of the animal responsible for eating another animal is a complicated topic that has been studied in detail over the years. With the advent of new 3D technologies, as well as the appearance of more advanced statistical approaches, many methods are currently available for the identification of carnivore activities based on their feeding patterns. The present study takes a look at the tooth marks left by tigers and leopards on bone, paying particular attention to the possible differences that may be product of sexual dimorphism. Here tooth pits produced by both male and female tigers and leopards have been studied using geometric morphometrics, revealing the sex of the animal to not be an important conditioning factor in tooth mark shape and size. Furthermore, both leopards and tigers are easily differentiable based on their tooth pits.

Abstract

Recent studies using geometric morphometrics for taphonomy have yielded interesting results, opening new horizons of research in both archaeological and paleontological sites. Here we present the analysis of tooth pits left by male and female individuals of two different carnivore species (Panthera tigris and Panthera pardus) in order to see if sexual dimorphism influences the morphology of tooth pit marks. In the process, 3D-scanning and applied statistics were used. Based on samples derived from two individuals of different sexes, the present results indicate sexual dimorphism in these felid species to not be a conditioning factor of tooth pit morphology.

1. Introduction

In recent years, taphonomic studies applied to the identification and characterization of carnivore intervention and consumption of prey have been reaching a great impulse. Great methodological achievements are being accomplished from an archaeological-paleontological perspective [1,2], with possibilities for further applications in forensic analyses of livestock predation [3].

In archaeological and paleontological contexts, investigation into carnivore activity is typically based on the tooth marks these animals leave on bone. From this perspective, many authors have specialized in the metric study of these traces [4,5,6], however, the amount of overlap observed between samples leaves much to be desired when classifying precise carnivore agencies. In modern day ecological contexts, many studies have delved into discerning the carnivores responsible for livestock predation, with methods based on the traces carnivores may leave on skin, meat, or hair [7,8,9,10]. Nevertheless, in cases where other scavengers or external processes may have intervened, leaving only the skeletal remains of the carcass, these traces are no longer present.

With the integration of more advanced applications in 3D technology and statistics in archaeological and paleontological contexts, [11] and [1] were able to design Geometric Morphometric (GMM) methodological approaches for the characterization of carnivore tooth pits on bone. The application of GMM substituted the traditional descriptive methods for statistical ones allowing the study of size and shape variation, and also allowing for the visualization of resultant covariation in terms of warping and transformation grids, see [12,13]. The GMM has been used frequently in different disciplines in the last 20 years, both in paleontology [14,15], archeology [16,17] and taphonomy [1,2,3,11,18,19,20]. Aramendi et al. [1] were able to reveal new insights into the carnivore species that may have intervened in the consumption of hominin fossils dated over 1.7 Mya. Likewise, [2] were also able to identify the activity of felids through tooth marks found on a 1.3 Mya Paranthropus boisei fossil from the site of Bell’s Korongo (BK, Olduvai, Tanzania).

Since these advances, multiple efforts have been made to increase the resolution of these different approaches, with the added benefit of using computational learning approaches to process such information [3,18,19,20,21,22]. In some of these cases, studies have been able to reach a resolution of correctly identifying the carnivore producing tooth marks with up to >90% accuracy based on their shape and size. These surprising results are not only able to efficiently differentiate between hyenid, felid, canid and ursid tooth marks [21], but also discriminate between species of the same taxonomic family, such as wolves and dogs [3].

Within this line of research, however, a number of questions still remain regarding possible conditioning variables that may distort these results. From this perspective, intraspecific analyses are becoming increasingly important, such as (1) the comparison of tooth marks found on different sized animals [19]; (2) possible intraspecific variability influenced by multiple individuals [22]; (3) or the reliability of experimental samples obtained by animals in captivity or in the wild [22]. In each of these cases, these kinds of queries are valuable since almost all tooth mark-related research is carried out on specific groups of individuals, without taking into consideration the possible variations that may occur when sampling from other populations. In the case where intraspecific variability is a considerable factor, then extrapolation to larger studies is much more difficult. Likewise, most tooth mark samples are developed on preys of a certain size, while animals in captivity have been observed to present different behavioral attributes to wild animals [23,24,25], thus affecting their tooth mark morphologies [22].

In each of the aforementioned analyses, observations made by authors tend to agree that intraspecific variability from different populations are less likely to affect tooth pits [21,22], at least in the case of wolves. To continue with this line of research, however, it is important that these results be observed equally in other carnivore families, such as felids, to extrapolate information on a larger scale. A recent study by Toledo et al. [26] revealed metric differences in the bite of female and male wolves on dental wax, while noting metric differences mainly in the size of male and female wolf teeth. While other studies have not yet noticed whether sexual dimorphism has an effect on the tooth marks produced [21,22], the sex of these individuals have not been directly observed. The present study considers the possible differences that may be produced by sexual dimorphism, specifically among felid individuals. Felids are well known for their anatomical and physiological differences among males and females [27,28,29]. Under this premise, if sexual dimorphism does not affect the tooth mark morphologies produced by felid individuals, then carnivores with much smaller dimorphic traits are equally less likely to present differences. For this purpose, the present study has paid particular attention to the tooth marks produced by male and female leopards (Panthera pardus kotiya Deraniyagala, 1956) and tigers (Panthera tigris sondaica Temminck, 1844) (Figure 1), two animals with great sexual dimorphism [27,29,30,31].

Panthera tigris sondaica [30] is a subspecies of tiger that ranges between 100–140 kg for males, while females typically range between 75–110 kg [31]. This implies a marked sexual dimorphism in terms of size, weight, and cranial dimensions, with males having larger and robust skulls [28]. In the case of tigers, dimorphism oscillates between populations, with greater differences existing in tigers originating from areas of higher latitudes [32]. Among skull features, the most dimorphic part is the snout [28].

Panthera pardus kotiya is a subspecies of leopard that ranges between 29 kg for females and 56 kg for males. Despite this, some specimens can reach up to 100 kg, varying depending on the environment [27,31]. While skulls of both sexes are morphologically similar, they can present variable sizes [29]. Some scholars suggest that the differences in the craniodental dimension of both sexes, along with the size of the animal, could suggest different eating habits [33].

2. Materials and Methods

2.1. Sample

The characteristics of carnivores used for this research are provided in Figure 1. In this case, the male individual of Panthera tigris studied weighed 120.0 kg and the female 86.2 kg. With respect to Panthera pardus, the male, 43.0 kg, is slightly larger than the female, 39.9 kg.

A total of 200 carnivore tooth pits were included and compared in this study; 104 pits produced by tigers (Panthera tigris sondaica,, Figure 1A: Female = 50; Male = 54), and 96 by leopards (Panthera pardus kotiya, Figure 1B: Female = 44; Male = 52).

Both samples originated from the Biopark zoo located in Fuengirola, Málaga (https://www.bioparcfuengirola.es/animales/). Both samples were therefore obtained from carnivores in captivity. No animals were sacrificed solely for the purpose of this study, so all the ethical standards regarding the treatment of animals for experimental purposes were followed. The experimental protocols employed consisted of the collection of bovine ribs with tissue, meat and articulated with vertebrae provided to each of the carnivores during weekly feeding over the course of 2 months. To ensure that samples were not contaminated, female and male individuals were clearly separated during feeding, while the bones recovered were differentiated and labelled accordingly. The bones were cleaned by boiling the sample with water.

2.2. Methods

Structured Light Surface Scanning (SLS) was chosen for all digitization of carnivore tooth marks, using the DAVID SLS-3 located at the Archaeometry unit of the Centre of Assistance for Investigation (CAI) in the Complutense University (Madrid). Models were captured as meshed point clouds in .ply format (Figure 2), which could later be processed using different software for landmark collection.

The present study used a 3D 30-landmark model proposed by [19,20]. This model consists of 5 fixed Type II landmarks marking the maximal length, width, and depth of tooth marks, followed by 25 computational landmarks projected across the entirety of the tooth pit using a patch (Figure 3). Landmarking procedures employed the use of the free IDAV Landmark 3.0.0.6 editor software. Landmarks were then formatted into morphological files for further processing using the R programming language (v.4.0.).

2.3. Data Analysis

Landmark data was first standardized through a Generalized Procrustes Analysis (GPA) [34,35]. GPA is an effective means of extracting morphological data quantitatively, allowing for the characterization of shape and form patterns across different configurations (orientation, translation, and scaling) [36,37]. Before any further hypothesis testing, allometric analyses were first considered to evaluate the impact that the tooth pit size might have had on the morphological variance. To achieve this aim, the logarithm of centroid sizes was calculated and used to perform regression on shape variables [38]. In case shape–size relationships proved to be of importance, final superimpositions were performed excluding the scaling step of the GPA procedure (otherwise known as form).

After normalizing, landmark coordinates were processed using dimensionally reduction in the form of Principal Component Analysis (PCA). PCA is a common visualization technique used to assess patterns in morphological trends. Each of these graphs were coupled with the computation of Thin Plate Splines [36,39] to visualize and assess these variations across each PC score. The PC scores representing the highest morphological variability (<95%) were then extracted for further processing.

The degree of statistical differences and similarities between samples was assessed using Two-One Sided equivalency Tests (TOST). This test assesses the magnitude of equivalency between samples according to Cohen’s δ [40]. For parametric versions of TOST, Welch’s t-statistic was used [41]. In cases where non-parametric tests were needed, the Yuen trimmed robust t-statistic was used [42,43]. It is important to note that, contrary to many other analyses of variance, both variants of TOST consider the Null Hypothesis (H₀) to indicate differences between samples.

Results of statistical hypothesis tests were additionally evaluated considering values of p < 0.003 (i.e., p < 3σ) to be a robust indicator of notable differences between samples [21,22,44,45]. p-values accompanied by calculations of their False Positive Risk (FPR) [45] supported this. FPR values are used to assess the probability that an observed p-value is a false positive, otherwise known as a Type I statistical error [45]. For FPR calculations a prior probability of 0.5 was adopted, assuming complete randomness [45].

The R programming language (v.4.0) was used to perform all statistics.

According to [22], we excluded the tooth scores from this analysis because this type of tooth mark shows intraspecific variability, and additional analysis is needed to evaluate the resolution of this type of tooth mark between the carnivores in captivity.

3. Results

In

Table 1. Results from allometric testing for each of the comparisons and their correspondent FPR value. Res.² = Squared residuals. FPR = False Positive Risk. M = Male. F = Female.

Groups Compared	F	Res.2	p-Value	FPR
Tiger vs. Leopard	2.7	0.013	0.014	14.0%
Tiger (M vs. F)	1.5	0.014	0.183	45.7%
Leopard (M vs. F)	3.0	0.024	0.045	27.5%

, both samples of tigers and leopards are described by non-allometric patterns, with the closest values resembling possible notable differences occurring when comparing species. Interestingly, and contrary to what would typically be expected in species of high sexual dimorphism, notable shape–size relationships are absent when comparing male and female animals. This is especially evident in the case of tigers, an animal presenting considerable size differences.

Geometric Morphometric analyses in shape space produced a principal component analysis represented by 90 PC scores, of which the first 11 represent up to 95% of the total sample variability (Figure 4). As can be seen, great overlap can generally be seen among samples, especially when comparing sexes (Figure 5). In general, both leopard and lions appear to present relatively “long” and “thin” pits, with leopards presenting a slightly stronger tendency towards more oval-shaped pits (PC2 = 17.29% of variance). Nevertheless, TOST equivalency testing revealed notable similarities between most samples, with the greatest differences appearing when comparing between species (

Table 2. Equivalency testing values for the different comparisons and their corresponding False Positive Risk (FPR) values. M = Male. F = Female.

Groups Compared	p-Value	FPR
Tiger vs. Leopard	0.042 1	26.6%
Tiger (M vs. F)	0.003	4.5%
Leopard (M vs. F)	0.009	10.2%

¹ Test results using non-parametric testing.

).

4. Discussion and Conclusions

The results obtained highlight aspects and inferences that might be relevant not only for archeology but also for sub-disciplines associated with zoology and applied taphonomy.

To begin, we see how the TOST test has determined that differences are likely to exist between leopards and tigers. At least, according to the present data, the odds of tiger and leopard pits being similar are at most 1 to 2.8 [44], and thus, with a prior probability of 0.5, these odds can be calculated at 1 to 1.4. Considering how other studies have shown high likelihoods of being able to differentiate between large and small felid species [21], the present data are not surprising because [21] was able to differentiate the pits tooth of leopards, lions, and jaguars, and in this work, we have been able to differentiate the pits tooth of leopards and tigers. Alongside other observations made by authors differentiating between carnivore species [3,18,22,46], we could even propose a more optimistic prior probability of 0.75 to report the posterior likelihood of tiger and leopard tooth pits being similar at 1 to 2 [44]. In either case, the present study supports that differentiation between carnivore species is possible.

Regarding the differentiation by sex, TOST values for both intraspecific comparisons affirm that in this sample there are no clear intraspecific differences between male and female samples based on their tooth pits. It should be noted that given the lack of an allometry study in the sample, we are comparing all of them only using the shape variable, and not the form variable, as in previous studies [18,19,20,21,22]. Despite this, and even if these tests were to be performed on form variables, leopards (p = 0.009, FPR = 10.3%) and tigers (p = 0.003, FPR = 4.5%) can be calculated to be similar in the form space as well.

These results support that the intraspecific size differences of a carnivore, either due to the population variability of the species or due to their sexual dimorphism [27,29,31], do not seem to be a conditioning factor in tooth pit morphology. Furthermore, considering how this has been confirmed in species presenting large sexual dimorphism, the present study strongly implies that tooth marks made by other species, such as those from the canid and hyenid families, may be less likely to be influenced by this variable.

The present study builds on hypotheses proposed by [19,20,22], in that carnivore tooth pits appear to be the type of tooth mark less affected by intraspecific variability, captive stress, prey size, and now sexual dimorphism. In light of this, it can additionally be argued that the results obtained by [1,3,18,22] are reliable for the discerning of carnivore agencies. This is yet another piece to the puzzle that argues that the biomechanical and tooth size variability that are often present [20,27] are not likely to distort these results.

These are promising results that provide a stronger basis for the wider application of our methodological approaches in paleontological, archaeological, and forensic veterinary fields, as well as modern-day ecological studies.

However, the results presented in this work must be taken with caution, since they have been obtained from a small sample and therefore the sample should be expanded in future studies. Despite this limitation, it must be assessed that different carnivores do not show intraspecific variability in tooth pits. Wolves do not show differences in tooth pits between different populations, and neither do leopards and tigers.