On reconstructing Giraffa sivalensis, an extinct giraffid from the Siwalik Hills, India

History of G. sivalensis discovery

Cautley (1838) briefly described the discovery of a remarkable vertebra in the Siwalik Hills in India. He believed the specimen to be very similar to that of extant giraffes—a significant statement, because up until that time no other fossil Giraffa species were known. Falconer & Cautley (1843) subsequently named the species Camelopardalis sivalensis and assigned the fossil, which was to become the holotype (Badam, 1979), as a third cervical vertebra. However, Lydekker (1885) disputed this and proposed that the holotype was in actual fact a fifth cervical vertebra of a ‘very small individual’. Since Cautley’s discovery, other Giraffa-like fossils have also been found in Asia, Europe and Africa, subsequently leading to proposals for species such as G. priscilla, G. jumae, G. stillei, G. gracilis, G. pygmaea and G. punjabiensis. The references to these fossil specimens are extensive, incomplete and confusing as can be seen by the references to G. sivalensis alone in Table S1.

Geographic and stratigraphic distribution of fossils

Matthew (1929) placed the upper Siwalik deposits, where G. sivalensis fossils and nearly all Siwalik fauna discovered by early writers such a Falconer have been found (Lydekker, 1876), as part of the Pinjor zone (Gaur, Vasishat & Chopra, 1985; Akhtar et al., 1991; Nanda, 2002; Nanda, 2008; Bhatti, 2004). The Pinjor zone dates to roughly 2.58–0.6 million years ago, placing the fauna discovered in this site as originating during the late Pliocene/early Pleistocene (Nanda, 2008). The site of discovery of the holotype for G. sivalensis was presented by Falconer & Cautley (1843) only as ‘the Sewalik range to the west of the river Jumna’ (currently the Yamuna river). Although Spamer, Daeschler & Vostreys-Shapiro (1995) described the locality as ‘Siwalik Hills, near Hardwar, Uttar Pradesh’, this is unlikely as Hardwar is east of the Yamuna. We therefore believe the locality was probably in the vicinity of the current Shivalik fossil park, Saketi, Himachal Pradesh, India (Fig. 1).

Size estimates and controversy

Size estimates of G. sivalensis have been inadequate or contradictory. For example, it has been proposed that G. sivalensis was about ‘one third shorter’ with a neck about ‘one tenth more slender’ than extant giraffes (Falconer & Cautley, 1843), and that the holotype belonged to a very small individual (Lydekker, 1885), that it had the same sized cranium as extant giraffes but with a shorter neck (Lydekker, 1876), that it was a large species but smaller than extant giraffes (Bhatti, 2004, p. 155), that it was of comparable size to modern giraffes (Bhatti, 2004, p. 255), that it was larger than extant giraffes (Mitchell & Skinner, 2003) and that certain proportions of the species’ neck were larger than extant giraffes (Lydekker, 1876, p. 105). Additional fossil specimens originally thought to belong to a separate species, G. affinis (Falconer & Cautley, 1843), were subsequently shown to belong to G. sivalensis and are currently believed to indicate a larger individual of the species (Lydekker, 1876, p. 105; Bhatti, 2004, p. 140). Table 1 summarises previous size estimates for G. sivalensis.

In this paper we outline and clarify the relevant information about G. sivalensis and its remains. In addition, we have made new estimates of its size and shape.

Studied material and dimensions measured

All postcranial specimens assigned to G. sivalensis that were available at the Natural History Museum in London were studied. From these specimens, body and neck size estimates were calculated using giraffe ontogenetic or available interspecific allometric equations. The only vertebra measured was the holotype (OR39747, Fig. 2), a cervical which had been extensively described by Falconer & Cautley (1843). A caudal fragment of a ‘fourth’ cervical (OR39748; Lydekker, 1885), also described as a second cervical by Falconer (1845), as well as a caudal part of a ‘third’ cervical (OR39746; Lydekker, 1885) were missing from the Siwalik collection at the Natural History Museum. Dimensions were measured with a vernier calliper and included: vertebral body length, cranial vertebral body height, cranial vertebral body width, caudal vertebral body height, caudal vertebral body width and spinous process length (Fig. 2).

Additional postcranial specimens assigned to G. sivalensis held at the Natural History Museum include fragments of two humeri (OR39749 and OR17136; Figs. 3 and 4 respectively), a fragment of a radius/ulna (OR17130) and various fragments of metacarpi and phalanges. All metacarpal specimens except OR39750 were avoided due to the unclear numbering of specimens and deformation of the fossils. Measurements of the long bones included length, midshaft circumference and midshaft diameter in craniocaudal and transverse planes. The length and circumference measurements were done with a measuring tape, while the cross sectional diameters were done with a vernier calliper.

Because there is no complete G. sivalensis skeleton its shape needs to be inferred as analogous to the only other extant Giraffa: G. camelopardalis. One of the methods of inferring body size from a model animal or animals requires that regression equations in the form y = mx^b (Huxley, 1932) be constructed. These regression equations can be based on data from different species (interspecific allometry), within the growth phase of a single animal (ontogenetic allometry) or amongst adult animals of different size but within the same species (static allometry). We applied ontogenetic as well as interspecific allometric equations to predict body mass in this case.

Ontogenetic data were obtained from previous studies by the authors (Mitchell, Van Sittert & Skinner, 2009; Van Sittert, Skinner & Mitchell, 2010; Van Sittert, Skinner & Mitchell, 2015). These data were used to construct allometric equations to describe body mass or body dimensions. The dimensions used from ontogenetic vertebral data are summarised in Table 2. Interspecific regression equations were sourced from previously published work (Anderson, Hall-Martin & Russell, 1985; Roth, 1990; Scott, 1990; Campione & Evans, 2012). The dimensions measured for the long bone ontogenetic data are summarised in Table 3.

There are inherent problems associated with using dental measurements as body size predictors, especially when only a single tooth is used (Damuth, 1990; Fortelius, 1990; Janis, 1990). Nevertheless, we have estimated size from teeth originally measured by Falconer & Cautley (1843), even though these teeth were initially assigned to a new species G. affinis, a species that was eventually abandoned (Lydekker, 1883). Uncertainty regarding these teeth specimens persisted until recent times (Spamer, Daeschler & Vostreys-Shapiro, 1995). Teeth specimens described by authors other than Falconer and Cautley which are noted in Table S1 were not evaluated further as there was either uncertainty regarding the authors’ species association (Lydekker, 1876), or the teeth specimens were not necessarily collected in the vicinity or stratigraphical layer of fossils described by Falconer and Cautley (Lydekker, 1878), or because certain specimens were deciduous. Table 4 presents dental specimens as well as dimensions as measured by Falconer & Cautley (1843). Body masses were estimated from regression equations established by Damuth (1990).

Statistical analyses

Allometric equations were generated from bivariate data through ordinary least squares regression. To facilitate this, measurements were logarithmically transformed to base e prior to analyses. According to Warton et al. (2006), ordinary least squares regression is appropriate when one wishes to predict y from x, even when x contains measurement error, as long as the results are interpreted in the context of ‘predicting y from x measured with error’. It is worth noting that there is controversy regarding the practice of logarithmically transforming data in scaling studies (Packard, Boardman & Birchard, 2009; Packard, Boardman & Birchard, 2010; Cawley & Janacek, 2010; Packard, 2013). The main argument is whether error becomes larger as body mass increases (multiplicative error), in which case logarithmic transformation is appropriate, or whether there is no correlation between error and body mass, in which case logarithmic transformation is not appropriate (Glazier, 2013). The debate is ongoing and will not be reviewed here. In this study we selected the method of log-transformation of data as it enables more convenient comparison among similar datasets.

Because body dimensions (especially body masses) can be predicted by different equations and by different fossil specimens, the predictions need to be validated. If regression equations had reasonable power in estimating body mass in both extant giraffids (G. camelopardalis and O. johnstoni), then they were regarded as robust enough to extrapolate to G. sivalensis as well. Therefore, dimensions of 10 okapi skeletons were recorded in addition to data obtained from G. camelopardalis. The okapi skeletons were housed in various museums and were recorded as the opportunities presented themselves (Table 5). Adult okapi specimens were assumed to have weighed 250 kg, with a range of 200 kg–300 kg (Lindsey & Bennett, 1999; Stuart & Stuart, 2006). The mature okapi specimens were identified through additional data associated with each museum specimen as well as by the degree of fusion of the epiphyses. The robustness of giraffe ontogenetic as well as interspecific equations to predict body mass in both adult giraffes and adult okapis correctly were assessed through the percent prediction error, calculated according to Smith (1984) and Van Valkenburgh (1990): Observed value−Predicted value/Predicted value×100.

Assumptions made

One of the major assumptions of this study is that G. sivalensis dimensions can be modelled from G. camelopardalis ontogeny. Although it is unusual to model an animal from the ontogeny of a different species it is not unique (for example, Roth, 1990). In assigning the holotype to a specific vertebra, we also assumed that there would be broad similarity in shape between the cervical vertebrae of G. sivalensis and G. camelopardalis. Falconer & Cautley (1843) illustrated this assumption to be the case for many but not all features of the holotype vertebra.

Another assumption was that the specimens used came from the same Giraffa species. We tried to use only those specimens that were clearly attributable to the Plio-Pleistocene and to the vicinity of the holotype discovery (Fig. 1), to limit possible confusion with other Giraffa species like G. punjabiensis. However, in some instances these criteria were not clear due to the lack of other samples or information, as in the discussion of vertebrae OR39746 and OR39748. Lastly, in terms of estimating body proportions in adult animals based on vertebral length, similarity in shape to G. camelopardalis was assumed.

Dimensions measured

The OR39747 and long bone dimensions measured are summarised in Tables 2 and 3 respectively, and where applicable the dimensions contain the equivalent measured values according to Falconer & Cautley (1843). Except for the cranial vertebral body height, our measurements on OR39747 are within 1%–5% of that reported by Falconer & Cautley (1843). Dimensions measured from okapi skeletons are presented in Table 5. Table 5 also contains predictions and prediction errors for okapi vertebral neck length based on G. camelopardalis ontogenetic data.

Predictions based on vertebra OR39747

Based on G. camelopardalis ontogenetic data, the average of dorsal and ventral neck length including soft tissue in G. sivalensis was 1,467 mm (y = 1.55x^0.859), the vertebral neck length excluding soft tissue was 1,270 mm (y = 10.66x^0.902) and the foreleg (hoof to withers) height in the living G. sivalensis adult was 2,540 mm (y = 7.61x^0.663, Table 6). This would mean that the reaching height of G. sivalensis was around 3.9 m.

The different vertebral dimensions predict the body mass to be within a range of 228 kg–575 kg, with an average of 373 kg (Table 7, 95% confidence interval (CI) ± 168 kg). We identified which of these dimensions could predict body mass accurately across species by calculating prediction errors when applying the G. camelopardalis regression equations to both extant giraffes and okapis. Naturally, because the predictions were done using G. camelopardalis ontogenetic allometry, the G. camelopardalis prediction errors were lowest (8%–50%). Predictions for okapi body mass, however, ranged from 17% to 99%. The only variable which provided relatively low body mass prediction errors in both okapi (17%) and G. camelopardalis (25%) was the caudal vertebral body dorsoventral height. This dimension predicts a body mass of 390 kg in G. sivalensis if we consider OR39747 as a third cervical. If OR39747 was considered a fourth or fifth cervical, body mass predictions will be 274 kg (y = 0.0011x^3.128) or 187 kg (y = 0.0004x^3.285) respectively (Table 7).

Predictions based on long bone dimensions

All of the G. sivalensis long bone specimens available at the Natural History Museum were incomplete proximally and/or distally. It was clear, nevertheless, that the bones had a similar slender appearance of extant giraffes and were elongated. Humeral specimen OR39749 was almost complete except for the proximal metaphysis, which has clearly broken off at the physeal line of a subadult animal. Regarding the radius/ulna specimen, the bones’ fusion at the midshaft was not complete as in modern giraffes, where the two bones are indistinguishable at midshaft in adults. The metacarpus specimen included in the study had the same caudal ‘columns’ or caudal groove as those evident in the extant giraffe (Solounias, 1999; Van Schalkwyk, 2004; Van Schalkwyk, Skinner & Mitchell, 2004) as well as in those of the okapi (own observation).

As no bones were complete lengthwise, bone length could not be used as a predictor for body mass which, in any case, has been shown to be a poor estimator of body mass in other taxa (Scott, 1990). Based on circumferences of the humeri (OR39749 and OR17136) and using G. camelopardalis ontogenetic data these specimens may have belonged to animals with body weight in the range of 770 kg–810 kg. An extant giraffe of this body mass would have a humerus length of about 477 mm–484 mm (y = 63.2(Bodymass)^0.304), which is just slightly longer than the 453 mm measured on OR39749 that lacked a distal metaphysis. The predictors based on radial and metacarpal cross sectional dimensions offered much higher body mass estimates, with averages of 1,024 kg and 1,107 kg respectively. In addition to employing ontogenetic data to generate allometric equations, we also referred to previously published interspecific studies (Anderson, Hall-Martin & Russell, 1985; Roth, 1990; Scott, 1990; Campione & Evans, 2012). Interspecific equations tended to predict heavier body masses than ontogenetic equations, especially so in the distal long bone samples.

Predictions based on dental dimensions

Four molars and two premolars were used for size predictions (Table 4), using equations developed by Damuth (1990). Body mass predictions based on tooth length (average = 577 kg, standard deviation = 155 kg) tended to be smaller than the predictions based on tooth width (average = 881 kg, standard deviation = 188 kg, t₍₂₇₎ = 4.83, p < 0.01). Predictions from molar length dimensions also tended to be lower than those from premolar lengths (average from molar lengths = 473 kg, average from premolar lengths = 682 kg, t₍₉₎ = − 3.12, p < 0.05).

Vertebral identity of OR39747

The anatomical identity of OR39747 was disputed by Lydekker (1885). He showed that Falconer was in a habit of not counting the atlas and axis as cervical vertebrae—which often meant that the start of the numbering of vertebra commenced at the third or occasionally the second postcranial vertebra. Mammalian C3–C5 forms a repetitive series and often does not have the distinguishing characteristics present in the other cervical vertebrae (Solounias, 1999). It is therefore indeed challenging to assign OR39747 to a specific vertebra. However, if we assume approximate similarity in shape between G. sivalensis and G. camelopardalis vertebrae, there are clues in the extent to which the cranial articular processes (Proc. articularis cranialis) extend beyond the body or centrum of the vertebra (Corpus vertebrae). In the G. camelopardalis C3, this process extends well beyond the cranial extremity of the vertebral body, but ends before or approximately at the same dorsoventral plane as the vertebral body in C4 and C5. Judging then by the extent of the articular processes of OR39747, it is a third, fourth or fifth cervical in decreasing order of likelihood. Falconer was therefore correct in assigning this vertebra as a third cervical, albeit fortuitously so.

Ontogenetic and interspecific scaling models

It is unusual although not unique to use ontogenetic allometry to predict an extinct animal’s size. For instance, Roth (1990) proposed that smaller animals of a species with distinctive morphologies (be they juvenile or adult) may still be better analogues than other taxa, at least in some aspects. We believe that this view is warranted in the current study as no extant species has such an extreme shape as G. camelopardalis. Predicting fossil masses from interspecific equations are further complicated by the decision of which taxa to include in regressions. For example, it is not clear whether predictions generated from interspecific allometric data are more accurate when based on closely related taxa with similar locomotor habits (Runestad, 1994; Janis, Theodor & Boisvert, 2002) or when using a wider sampling base (De Esteban-Trivigno, Mendoza & De Renzi, 2008). Other factors that may influence precision of body mass predictions in interspecific studies are body mass estimations (instead of body mass measurements), small intrataxa sample sizes, and overrepresentation of animals of one sex or of exaggerated proportions. To overcome these problems, we investigated which ontogenetic scaling parameters, if any, might be suitable and robust enough for predictions amongst extant Giraffidae. It is possible that giraffe ontogenetic equations are also acceptable for comparison not just amongst the Giraffidae but amongst, for example, extant camelids with similar gaits. However, okapis were considered as an adequate reference in this case as they are closest to giraffes phylogenetically and because the ontogenetic scaling of their long bones scales differently to other cetartiodactyla (Kilbourne & Makovicky, 2012).

Ontogenetic scaling and interspecific scaling exponents are generally not interchangeable (Gould, 1966; Pélabon et al., 2013); in this case it is dependent on the assumption that G. sivalensis had a similar body plan as juvenile extant giraffes. We thus found it appropriate, where possible, to test both ontogenetic and interspecific curves to infer proportions of G. sivalensis, but realised that neither of these methods may be appropriate for each and every dimension measured.

Neck length and reaching height

Badlangana, Adams & Manger (2009) presented interspecific predictions for vertebral neck length based on vertebral body length. Using their data (presented in Table 1 of Badlangana, Adams & Manger, 2009), we could estimate G. sivalensis C2–C7 vertebral neck length as 1,150 mm (Table 6), slightly shorter (45 mm or 4%) than vertebral neck length calculated from our ontogenetic data. There are therefore reasonable grounds to believe that our estimated neck length based on ontogenetic data is valid, or at least close to interspecific curves. Further support for this rationale can be seen where the G. camelopardalis ontogenetic curve gives appropriate predictions for vertebral neck length in both the extant giraffe and okapi (Fig. 5). Extant adult giraffes have an average external neck length of about 2,013 mm in males (1,000 kg and above) and 1,832 mm in females (800 kg and above) (Mitchell, Van Sittert & Skinner, 2009). Assuming the same body plan for G. sivalensis as for G. camelopardalis, then G. sivalensis had around 350 mm (20%) to 550 mm (27%) shorter necks than modern giraffes, depending on whether OR39747 came from a female or male animal. This is a slightly longer neck length than Falconer & Cautley’s (1843) estimated neck length for G. sivalensis, which is approximately a third shorter than extant giraffes.

Body mass

The body mass predictions for G. sivalensis are wide (Fig. 6). Possible reasons for the large range of predictions are that certain fossils were erroneously attributed to G. sivalensis and/or that certain specimens and allometric equations are inadequate for body mass predictions. Before decisions could be made regarding the validity of attributing a fossil to G. sivalensis, we ascertained the equations that were robust enough to predict body mass accurately across species.

Vertebra OR39747 body mass estimates

It is unconventional to use vertebrae as proxies for body mass, although due to the lack of other samples it has been done before (see for instance Taylor, 2007; Taylor & Naish, 2007). As OR39747 is the holotype, it necessitates that body mass estimates are made from it if other Giraffa spp. specimens are to be attributed to it. Although vertebral body length has higher R² values than cross sectional vertebral properties (Table 7, Van Sittert, Skinner & Mitchell, 2010), cross sectional properties are still preferable predictors of body mass in this case. The first reason is that R² value is inferior to percent prediction error (% PE) and percent standard error of the estimate when assessing reliability of body mass predictions through regressions (Smith, 1984). Secondly, vertebral cross sectional properties are subjected to the stresses and strains within the neck (Slijper, 1946) and therefore are a much better indicator of head and neck mass and by implication body mass. Conversely, vertebral body length is influenced by factors other than body mass such as the number of vertebrae in an anatomical area (compare birds and mammals’ cervical region) or the lifestyle of the animal. We found caudal vertebral height (dorsoventral diameter) to have the lowest % PE (25% and 17%) when predicting body mass in both extant giraffes and okapis respectively (Table 7 and Fig. 7), and therefore considered this dimension to be most robust for body mass predictions across giraffids. There are no other published interspecific regression equations using vertebral dimensions for the prediction of body mass in ungulates of which we are aware. The caudal vertebral height predicts a body mass of 390 kg in G. sivalensis. Interestingly, the average body mass prediction from the remaining vertebral regression equations (C3 vertebral body length, cranial height, cranial width and caudal width, Table 7) is fairly similar—368 kg. The only body mass prediction to fall outside the 95% confidence interval based on all vertebral dimensions including vertebral height (373 kg ± 119 kg) is vertebral body length, predicting a mass of 575 kg.

Nevertheless, the body mass prediction from caudal vertebral height could be either an over or underestimate. Considering it as an overestimate would mean that this animal had a relatively heavy neck and head complex but a slender or lightweight body. This is unlikely as a larger head and neck complex is unsupportable unless accompanied by a larger total body size (Taylor & Wedel, 2013). Conversely an underestimate would mean a slender neck and head complex but a relatively stocky body. This is a more plausible scenario and if indeed it is the case, it might explain the discrepancy between vertebral and dental body mass predictions when compared to those of limb bones.

Limb bone body mass estimates

Interspecific long bone cross sectional properties, although probably more closely related to body mass than any other variable, have nevertheless been found to be poor predictors of body mass in giraffes and in some cases, okapis (McMahon, 1975; Anderson, Hall-Martin & Russell, 1985; Scott, 1990; Janis, Theodor & Boisvert, 2002), although it should be noted that a recent interspecific study has shown giraffes to be more amenable to interspecific equation predictions (Campione & Evans, 2012). Similarly, we found higher prediction errors with interspecific equations compared to G. camelopardalis ontogenetic curves, with a 5% prediction error based on humeral ontogenetic data (Fig. 7). Errors were inflated when using more distal bones. Therefore, the most appropriate long bone variable useful for G. sivalensis body mass determination is very likely humeral cross sectional properties, using our ontogenetic G. camelopardalis sample.

The average body mass estimated from humeral ontogenetic analysis is 732 kg. Interestingly, this body mass is about 150 kg more than would be indicated by a G. camelopardalis of similar neck length, and 342 kg more than the mass predicted from OR39747 cross sectional properties. This could mean that either the humeral fossil specimens were incorrectly assigned to G. sivalensis, that G. sivalensis had a relatively stockier body and thinner neck than G. camelopardalis or that the holotype vertebra came from a female and the humeral specimens from large males.

Unfortunately, none of the other long bone dimensions seem to be reliable predictors of body mass across extant giraffids. The best non-humerus candidate using interspecific scaling seems to be the radius transverse diameter with a 43% and 11% prediction error in okapis and giraffes respectively. This dimension predicts that the specimen belonged to an animal of approximately 1,238 kg, which suggests this animal might have been heavier than G. sivalensis. There are no interspecific equations for metacarpi that we could find, and therefore we could only rely on ontogenetic equations. Yet, similar to the radial prediction, the metacarpal transverse diameter predicts a body mass of 1,165 with around 20% prediction error. The inflated prediction errors could be because humeri and femora are generally more suitable for body mass predictions than more distal bones, especially in giraffes (McMahon, 1975). It is also possible that the fossil long bones were incorrectly assigned to G. sivalensis and perhaps belonged to another similar species existing at the same time and location.

Dental body mass estimates

There have been numerous dental specimens ascribed to G. sivalensis (Table S1). Unfortunately, not all of these specimens are from the same locality and are probably from different stratigraphic zones. Subsequently, there appeared to be uncertainty regarding the correct species allocation of these fossils (see especially Lydekker, 1876). A discussion on the morphology and correct species classification of teeth specimens assigned to G. sivalensis were not considered as part of this study, and we therefore used only those teeth mentioned by Falconer & Cautley (1843). These specimens were originally assigned to the species G. affinis—a classification later abandoned by Falconer himself and also disputed by Lydekker (1883), who re-assigned the fossils to G. sivalensis. As the specimens originated from the same area and strata as the holotype OR39747, which is the Pliocene of the Siwaliks (Lydekker, 1885), we believe it reasonable to consider them as truly G. sivalensis teeth until further evidence emerges.

Molar length measurements are more reliable indicators of body mass than molar width or area (Damuth, 1990; Fortelius, 1990; Janis, 1990). Furthermore, Janis (1990) found that premolar row length are poorer correlates than molar row length. Molar lengths predict an animal within the range of 288 kg–673 kg, which is similar to OR39747’s caudal vertebral height body mass prediction of 390 kg.

Combined size estimates

Lydekker’s (1885) suggestion that OR39747 belonged to a small individual could have meant that the animal was still immature, that the animal was a relatively small individual of the species or that the species itself was small within the genus. It is unlikely that Lydekker meant an immature animal as the fusion of the epiphyses to the body of the vertebra is complete and clear definitions of bony ridges and muscular depressions indicate a mature animal (Falconer & Cautley, 1843). Lydekker might have based his idea of a small individual on two larger vertebrae assigned to G. sivalensis—a proximal part of a ‘third’ and distal part of a ‘fourth’ cervical, OR39746 and OR39748 respectively (Lydekker, 1885, Table S1). Unfortunately, these vertebrae were not locatable within the Siwalik collection at the time of this study (P Brewer, pers. comm., 2013, Curator of fossil mammals, Natural History Museum), and we subsequently could not measure them. Nevertheless, Falconer (1845) reported OR39748 to be 2.1 inch (53.3 mm) in height and width at the caudal extremity, which is only 0.2 mm greater and 0.1 mm less than our respective measurements of OR39747 (Table 2). Based on ontogenetic allometry for caudal vertebral body height, OR39748 came from an animal weighing 394 kg or 277 kg, depending on whether it was a C3 or C4 vertebra respectively (Table 7). The animal from which the holotype vertebrae originated was therefore also not relatively small compared to the size estimated from specimen OR39748. It is possible though, especially considering body mass estimates from the humerus, that there might have been sexual size dimorphism present in G. sivalensis. If that is indeed the case, OR39747 and OR39748 would have been females about half the size of fully grown male animals, a possibility also supported by the fossil teeth considered in body mass estimates.