Cautionary tales on the use of proxies to estimate body size and form of extinct animals

2.1. Dunkleosteus terrelli (Placodermi: Arthrodira)

Dunkleosteus terrelli is a large late Devonian arthrodire placoderm, known for its bony armour and guillotine‐like jaws. However, outside of its bony head and trunk armour (Figure 2a), the rest of its body was composed of perichondrally ossified cartilage. This material rarely preserves, complicating estimates of body size and morphology (Carr, 2010; Ferrón et al., 2017). Additionally, Dunkleosteus has no close living relatives that can be used to interpret its anatomy. Complete remains are known for some smaller arthrodires (Jobbins et al., 2022; Miles & Westoll, 1968), but these taxa likely differed from D. terrelli in ecological niche and body shape (Ferrón et al., 2017).

Dunkleosteus has been traditionally cited as reaching 5–10 m in total length (TL). However, these are generally speculative estimates that omit key methodological details as to how they were produced, including what taxa or anatomical elements were used as proxies, what specimens of Dunkleosteus were examined and the measurements used to produce these estimates (see Engelman, 2023b; Ferrón et al., 2017 for details), making these estimates non‐reproducible and highly questionable. One exception was Ferrón et al. (2017), which calculated the TL of D. terrelli using the upper jaw perimeter of select large, nektonic extant sharks to estimate its caudal fin shape, as their method required body size as an independent predictor variable. This study assumed mouth and body size likely correlate in fishes because predator size is correlated with prey size, which in turn correlated with gape, meaning mouth size and body size may be indirectly linked (Ferrón et al., 2017). However, this study did not test if sharks and arthrodires showed comparable mouth proportions or cross‐examine the accuracy of the resulting body length estimates (Engelman, 2023b; Ferrón et al., 2017). Testing this method on complete arthrodires finds arthrodires have proportionally larger mouths than sharks, producing overestimates of TL 2–2.5 times the actual value (Engelman, 2023a, 2023b). Traditionally cited lengths for Dunkleosteus also require a hyper‐elongate trunk, with a head of only ~8% TL (see Figure 1), and head–body proportions that are more extreme than even eels. Not only are these proportions implausible, but they are also unlike complete arthrodires or non‐anguilliform fishes (in which head length is generally ~17–30% TL; Engelman, 2023b).

Because of the inability of mouth dimensions to reliably estimate the size of arthrodires, Engelman (2023a) attempted to estimate the size of Dunkleosteus based on the combined length of the neurocranial and branchial regions of the head (i.e., head length minus preorbital length; Engelman, 2023a), reasoning that the scaling of these areas with body size was likely strongly constrained due to their function. This study (Engelman, 2023a) selected a variable based on homologous landmarks between Dunkleosteus and a wide range of extant proxies. It also tested the initial assumption of correlation to ensure that scaling patterns were consistent across taxa. Furthermore, it used complete arthrodires as case studies to ensure that the method accurately predicted body size in this group, suggesting that this method should produce reliable sizes for Dunkleosteus in the absence of complete specimens. This resulted in lengths of only 3.4–4.1 m for large D. terrelli individuals, with even the upper ends of the margin of error barely overlapping with the smallest lengths in previous studies (Engelman, 2023a). Independent tests by scaling from other arthrodires produced similar results, and the reconstructed body shape for Dunkleosteus better agrees with the comparative anatomy of other arthrodires (e.g., in fin location/size relative to TL and trunk armour dimensions; Engelman, 2023a, 2024). Arthrodires, actinopterygians and chondrichthyans showed significant, clade‐specific differences in overall body form, with complete arthrodires exhibiting a much stockier body plan compared to chondrichthyans (Engelman, 2023a).

2.2. Helicoprion spp.: (Chondrichthyes: Eugeneodontidiformes)

Helicoprion, found globally in Permian deposits, is arguably the most widely recognised Palaeozoic chondrichthyan. Its notoriety stems from the unique nature of its spiraliform tooth whorls (Figure 2b), which constitute practically the only known remains of the taxon. The lack of any reasonable extant or extinct proxies to the tooth whorls of Helicoprion has presented significant difficulties in inferring its body size and form. For some time after the description of Helicoprion by Karpinsky (1899), there was no agreement as to the position of the tooth whorl on the body of the animal or even its nature (e.g., dermal spine or oral teeth), making inference of body size or form impossible. Although Karpinsky (1899) considered other interpretations, he favoured a location of the tooth whorl mostly external to the mouth and extended from the upper jaw. However, based on examination of IMNH (Idaho Museum of Natural History) 37899, a unique specimen of H. davisii from the Permian Phosphoria Formation of Idaho, USA, with remains of cranial cartilage preserved, Bendix‐Almgreen (1966) concluded the tooth whorl was located at the symphysis of the lower jaw, internal to the buccal cavity, and that no similar tooth whorl was in the upper jaw.

Subsequently, Lebedev (2009) made the first quantitative reconstruction of the body size and shape of Helicoprion, using (1) extant odontocetes (toothed whales) and (2) extinct eugeneodonts known from nearly complete remains (Caseodus, Romerodus, and Fadenia) as proxies. The first proxy was based on a presumed similarity in size, diet (fish and cephalopods) and habitat (pelagic) and the second on phylogenetic proximity. Lebedev's reconstruction featured an elongated lower jaw, with the tooth whorl placed near the distal end, and conventional fusiform fish body proportions. Based on the eugeneodont proxies, Lebedev inferred a head length of 2.5–4.0 times the tooth whorl diameter. An assumption of a ratio of body length to head length of 5.0 (conventional fish body proportions) yields a body length (TL) of 12.5–20 times the tooth‐whorl diameter. For a tooth whorl with a diameter of 0.56 m (the largest one listed in Tapanila & Pruitt, 2013), this corresponds to a range of 7.0–11.2 m TL. Later examination by Tapanila et al. (2013) of IMNH 37899 by X‐CT (X‐ray computed tomography) showed that Helicoprion's lower jaw was shorter than Lebedev envisioned and that the ratio of the head length to the tooth‐whorl diameter was approximately 2.5 (Tapanila et al., 2020, figure 3), resulting in an inferred body length of approximately 7.0 m TL.

2.3. Otodus megalodon (Chondrichthyes: Lamniformes)

The megatooth shark, Otodus megalodon, is an iconic extinct elasmobranch represented primarily by gigantic teeth (Figure 2c) commonly found from the mid‐Miocene to the Early Pliocene nearly worldwide (Boessenecker et al., 2019; Cappetta, 2012; Cooper et al., 2020, 2022; Gottfried et al., 1996; Sternes et al., 2023). The biology of O. megalodon has remained difficult to decipher. Although some vertebral remains, placoid scales, and fragments of tessellated cartilage of O. megalodon have been reported (Bendix‐Almgreen, 1983; Leriche, 1926; Shimada et al., 2023), the limited fossil record has hampered attempts to estimate its true size and to reconstruct body form. When such attempts are made, the extant white shark (Carcharodon carcharias) has been assumed to be a logical ecological proxy due to its similar tooth form and trophic ecology to O. megalodon (Collareta et al., 2017; Gottfried et al., 1996).

The most common size estimation method for O. megalodon has been extrapolation from scaling relationships between dental measurements and total body lengths in C. carcharias (Perez et al., 2021; Randall, 1973; Shimada, 2002, 2019). These studies generally suggest a maximum length of 15–20 m for O. megalodon. Alternatively, the allometric relationship of vertebral diameters in extant C. carcharias (Gottfried et al., 1996) has been used to extrapolate body size based on incremental growth bands preserved in a set of vertebral specimens of O. megalodon, producing a length estimate at birth of 2 m (Shimada et al., 2021).

Likewise, the body form of O. megalodon has traditionally been modelled after extant C. carcharias (Bendix‐Almgreen, 1982; Cooper et al., 2020, 2022; Gottfried et al., 1996). This seemed logical especially in earlier studies (Applegate & Espinosa‐Arrubarrena, 1996; Gottfried et al., 1996), due to the similarly serrated dental morphology, the fossil shark was placed in the genus Carcharodon (Lamnidae) with the interpretation that ‘C.’ megalodon was the direct ancestor of extant C. carcharias. As of now, the fossil species is considered to belong to Otodus (Otodontidae) rather than Carcharodon (Shimada et al., 2017). Despite C. carcharias and O. megalodon likely not having any direct phylogenetic link within Lamniformes (Shimada, 2022; Shimada et al., 2017; Sternes et al., 2023, 2024), the use of C. carcharias as a proxy has continued based on the assumption that O. megalodon was an active, regionally endothermic (Ferrón, 2017) predator like the extant lamnids (Cooper et al., 2020, 2022).

Geochemical evidence has subsequently confirmed that O. megalodon was a ‘warm‐blooded’ shark (Griffiths et al., 2023) occupying a high trophic position (Kast et al., 2022; McCormack et al., 2022). However, the assumption that O. megalodon must have resembled extant C. carcharias has been called into question. Support for the relationship between body form and thermophysiology in sharks has weakened (Dolton, Jackson, et al., 2023; Dolton, Snelling, et al., 2023; Sternes et al., 2023), and the absence of keels on its scales suggests that O. megalodon was likely not a fast swimmer like extant lamnids. Additionally, the total combined length of a partial vertebral column of O. megalodon from the Miocene of Belgium, previously estimated to be from a 9.2‐m‐TL individual based on comparisons of vertebral diameters in extant C. carcharias (Gottfried et al., 1996), was measured to be 11.1 m (Cooper et al., 2022). This discrepancy – that is, the actual measured length of the incomplete vertebral column (not considering the head or caudal fin) being substantially longer than the TL estimate for this specimen based on the extant C. carcharias (9.2 m TL) – strongly indicates that O. megalodon was not merely a larger version of extant C. carcharias but rather a proportionately slender, more elongate shark (Sternes et al., 2024). The exact body morphology and maximum TL of O. megalodon remain uncertain, but the implications are significant. Not only may the use of extant C. carcharias not be appropriate for deciphering the body form and body size of O. megalodon, but the observation also strongly indicates that all previous TL estimates based on dental proportions of extant C. carcharias, which concluded a maximum TL of at least 15 m for the species (Shimada, 2019), may be underestimates.

2.4. Perucetus colossus (Mammalia: Artiodactyla)

Perucetus colossus is a recently described, large‐bodied archaeocete (stem cetacean) from the late middle Eocene of Peru (Bianucci et al., 2023). This taxon is only represented by the holotype, consisting of 13 vertebrae (Figure 2d), 4 ribs, and the right innominate. Perucetus is considered a member of Basilosauridae, the derived group of fully aquatic archaeocetes that may represent the sister group of crown Cetacea. Perucetus exhibits extreme pachyostosis and osteosclerosis, traits generally associated with shallow diving in modern aquatic mammals (Buffrénil et al., 2010).

Despite belonging to a clade with modern representatives (crown Cetacea), there are no close extant analogues for the large, serpentiform body plan of basilosaurids. Thus, to reconstruct the body size of Perucetus, Bianucci et al. (2023) utilised multiple proxies including sirenians and neocetes. Bianucci et al. (2023) first scaled up and dilated a digital 3D model of the skeleton of the basilosaurid Cynthiacetus peruvianus such that the volume of its bones equalled that of the corresponding skeletal elements of P. colossus (producing multiple models to account for the uncertain position of the latter's vertebrae and variable vertebral count among basilosaurids), yielding estimates of skeletal length and volume ranging between ~17–20 m and 2.9–4.1 m³, respectively. Body mass values were then calculated for the P. colossus holotype based on the range of skeletal fraction (SF) observed in extant marine mammals, yielding minimum and maximum estimates of 85 t (assuming the mean SF of large Trichechus manatus and the minimum skeletal volume) and 340 t (assuming the SF of Mesoplodon europaeus and the maximum skeletal volume), respectively (Bianucci et al., 2023).

Subsequently, Motani and Pyenson (2024) raised several criticisms of Bianucci et al. (2023)'s analysis. For example, Motani and Pyenson (2024) suggested that it would be impossible to fit 180–340 t of biomass (corresponding to the upper portion of Bianucci et al. (2023)'s range of body mass estimates) into the volume of a 20 m whale, with their volumetric methods instead producing estimates of 60–114 t using volumetric methods. They also questioned Bianucci et al. (2023)'s use of a simple SF ratio as opposed to an allometric regression equation, arguing that across marine and terrestrial mammals (including both neocetes and sirenians), SF appears to scale with positive allometry rather than isometry. Using regression equations, Motani and Pyenson (2024) produced estimates of 135–193 t assuming a neocete‐like SF and 40.0–54.6 t using a sirenian‐like SF (Motani & Pyenson, 2024). Nevertheless, this revised body mass estimate is not free of limitations either, namely, reliance on a model created from a 2D palaeoartistic reconstruction that was not depicted in direct lateral view. This method is by no means intrinsically flawed but is heavily dependent on approximation of three‐dimensional body proportions. Alternatively, a silhouette could have been produced from the modified Cynthiacetus model presented by Bianucci et al. (2023), which could have yielded a different result.

4.1. Modelling and extrapolation

The most fundamental requirement for predicting body size and/or form of an extinct species is that the proxy accurately predicts body size in the first place (Bates et al., 2009). This might seem trivial, but different physical features exhibit diverse scaling relationships across taxa, and the selection of specific anatomical units to model body size requires careful examination and abundant data (e.g., Field et al., 2013). Nelson et al. (2023) found that limb bone cross‐sectional dimensions (diameters and circumferences), long considered the strongest predictors of body size among terrestrial vertebrates (Anderson et al., 1985; Campione & Evans, 2012; Ruff, 1990), had significant, non‐random bias independent of phylogeny but seemingly correlated with body robustness – a factor that is difficult to control for mathematically without circular logic. Unfortunately, existing studies often do not explore possible sources of allometric bias or uncertainty when selecting size proxies, but assume a close and consistent relationship with body size between the proxy and focal taxa a priori (Cooper et al., 2020, 2022; Ferrón et al., 2017; Gottfried et al., 1996). This issue is exemplified by initial body size estimates for Dunkleosteus, which were based on scaling relationships of the upper jaw in extant sharks (Ferrón et al., 2017) despite a lack of evidence for similar jaw proportions between arthrodires and sharks (Engelman, 2023b). While post hoc analysis of arthrodire cranial scaling improved size estimates for Dunkleosteus (Engelman, 2023a, 2023b), it is extremely challenging if not impossible to verify the validity of proposed scaling trends in other taxa with few or no living (or well‐preserved fossil) relatives, such as Helicoprion. Whenever possible, proposed scaling relationships should be tested empirically using close relatives of the focal taxon to ensure model reliability (Engelman, 2023b) or cross‐testing size/form estimates using multiple proxies and/or scaling methods (Bianucci et al., 2023; Motani & Pyenson, 2024). Where this is not possible, we urge scepticism of resulting body size and form estimates.

Another common issue is extrapolation error. When using regression equations to estimate body size, imprecision typically increases as the size of anatomical elements from the focal taxon extends beyond that of the training data (Bates et al., 2009; Engelman, 2023a; Schmidt‐Nielsen, 1984). This is a particular problem if taxa of interest to palaeontologists are significantly larger or smaller than their nearest extant proxies or occupy size ranges where only a limited number of similar‐sized modern proxies exist (i.e., megafauna). This can be mitigated if taxa in the training data span a wide range of body sizes (Campione, 2017) but such wide samples are often not available. Perhaps, the best example of the effects of extrapolation error on body mass estimates can be seen in the extinct rodent Josephoartigasia (Engelman, 2023a; Millien, 2008), although it also applies to other taxa, including Perucetus (Bianucci et al., 2023; Motani & Pyenson, 2024).

Many size estimates run into problems with the ever‐present spectre of positive or negative allometry (Schmidt‐Nielsen, 1984), either assuming their anatomical proxies scale isometrically with body size or calculating their estimates via simple scaling ratios with a proxy taxon (which implicitly assumes isometry). However, isometry is typically the exception, not the rule, among scaling relationships (Raup & Stanley, 1978: p. 61). Regression models tend to be more robust to this kind of bias due to their variable slope, whereas simple scaling ratios can be biased by even slight deviations from isometry or if the size proxy chosen does not show a strong correlation with body size (Grillo & Delcourt, 2017). As with extrapolation error, sampling a wide array of body sizes in the training dataset is one of the best ways to detect positive or negative allometry. Body size estimates should always be made based on regression equations or volumetric models whenever possible. Estimating body size via simple scaling ratios from one or a few proxy specimens or taxa should be considered as a last resort if appropriate regression models are not available, and the resulting size and/or form estimates considered very tentative until more robust estimates of size can be produced.

Other issues arise from the fact that most variables in allometric scaling relationships are logarithmically distributed and thus log‐transformed before analyses. It is often assumed that relationships between variables are completely linearised by log transformation (Engelman, 2022b; Schmidt‐Nielsen, 1984). However, this is not always the case, and further inspection has shown some biological relationships previously thought to be log‐linear may, in fact, scale log‐curvilinearly (Bertram & Biewener, 1990; Engelman, 2022a; Knell, 2009; Venditti et al., 2024) with their curvature implying treating these variables log‐linearly may overestimate body size. This is problematic as the biological significance of log‐curvilinear relationships is not well understood, nor do their mathematical constants have a ready explanation (Knell, 2009; Manger et al., 1999) in contrast to log‐linear models which follow a power law (Schmidt‐Nielsen, 1984). Until we have an improved understanding of non‐linear allometry, the validity of linear approximations to these allometric relationships cannot be known. For this reason, we suggest future studies at least consider the possibility of non‐linear allometry in their datasets and report suspected log‐curvilinear relationships if found.

Log‐transformation creates other issues in model evaluation and prediction. Most regression models measure the strength of correlations between variables using the coefficient of determination (r ²), but for log‐transformed models (especially ones intended for prediction), r ² is actually a poor measure of relationship strength. Coefficients of determination tend to unilaterally increase as data spread increases, and log‐transformation exacerbates this problem because it compresses the scatter of data points around the regression line (Smith, 1984; Van Valkenburgh, 1990), inflating r ² values. This means that even log‐scaled models with r ² greater than 0.9 can have poor prediction accuracy in practice (Smith, 1984; Van Valkenburgh, 1990). For this reason, percent error (%PE) and percent standard error of the estimate (%SEE) are often preferred as measures of predictive accuracy because they directly measure the accuracy of the predicted values (Campione & Evans, 2020; Engelman, 2022a, 2023a; Van Valkenburgh, 1990).

Log‐transformed regression equations also tend to produce unreasonably large prediction intervals, often on the scale of orders of magnitude. This is because antilog transformation turns the normally distributed residuals of a log‐scaled regression equation into non‐normally distributed (leptokurtotic) residuals on an arithmetic scale (Bates et al., 2009; Bertram & Biewener, 1990; Engelman, 2023a) with extremely long ‘tails’ to the resulting distribution. This in turn results in large error bars and substantial uncertainty in body size estimates, often beyond what is morphologically plausible. Phylogenetic comparative least squares and volumetric estimation methods provide possible mitigation measures, although in both cases reduced error range comes at a trade‐off with prediction accuracy (Campione, 2017; Campione & Evans, 2012). While this uncertainty is unavoidable, it should be accounted for when making biological inferences about palaeoecology. All mathematical models have assumptions and limitations, the validity of which should ideally be carefully considered when selecting proxies from which to estimate body size and/or form in extinct taxa.

4.2. Incomplete specimens

In most cases, the body size and/or form of extinct animals must be estimated from extant proxies due to the lack of complete fossil specimens. At a basic level, it is impossible to confidently and accurately predict the body size or form of an extinct organism without complete specimens as researchers have no way of knowing the size or shape of missing elements without relying on inference from proxy taxa. This issue affects each of the four case studies discussed here, none of which are known from complete specimens (Bendix‐Almgreen, 1966; Bianucci et al., 2023; Engelman, 2022b; Sternes et al., 2024; Figure 2). Indeed, the fragmentary nature of remains used to reconstruct the body size and form of many extinct taxa can substantially increase error. Fragmentary remains are often first used to estimate the size of some larger or complete morphological structure (e.g., skull), which is in turn used to approximate total length. These cascading assumptions result in the propagation of error at each stage of reconstruction (Molnar & Vasconcellos, 2016), further complicating downstream ecological, evolutionary and biomechanical interpretations.

A lack of anatomically complete specimens also increases the likelihood of existing remains being misinterpreted as belonging to different species and/or parts of the body. Errors in body size and form estimation resulting from anatomical misinterpretation or misdiagnosis can be seen in Helicoprion, where the consensus position of the tooth whorl on the body has changed on multiple occasions (Bendix‐Almgreen, 1966; Karpinsky, 1899). Similarly, extreme modifications in the available vertebrae of Perucetus relative to other basilosaurids prevent a precise identification of their position within the vertebral column, which adds considerable uncertainty to body size estimations. For Dunkleosteus, lengths of 5–10 m were at least partly based on a priori assumptions of this taxon exhibiting a greatly shortened trunk armour compared to other arthrodires, which were never validated and indeed subsequent observations showed it was likely incorrect (Engelman, 2024). Unfortunately, the limitations associated with incomplete fossil specimens are difficult to overcome without new palaeontological evidence. Yet, this issue highlights that body size and/or form estimates are intrinsically uncertain if lacking adequate fossil material. Wherever possible, studies should take this into consideration and acknowledge the potential for new palaeontological interpretations of the fossil specimens upon which estimations are based.

4.3. Intraspecific variation

Another important consideration when selecting proxies for estimating body size and form in extinct taxa is intraspecific variation. Ontogeny and sexual dimorphism exert substantial influence on both body size and shape (Hone et al., 2016; Mallon, 2017; Motani et al., 2018; Paiva et al., 2022; Sanchez‐Villagra, 2010), and this needs to be taken into account when reconstructing extinct species. Some studies use regression equations derived from allometric patterns within a single species to produce their size estimates (e.g., Crocodylus porosus and Gavialis gangeticus in Sereno et al., 2001; Physeter macrocephalus in Lambert et al., 2010; and Carcharodon carcharias in Gottfried et al., 1996), and this raises concerns about conflating intraspecific patterns of allometry across the growth curve of a single species with true patterns of interspecific allometry in mature individuals (Paiva et al., 2022). Several studies on extinct megafauna have noted that even very large individuals ‘still appear to be growing at the time of death’ based on sutural fusion and bone microstructure (Buchy et al., 2003; Evans et al., 2014; Hone et al., 2016; Lomax et al., 2024), yet in many cases still appear to be sexually mature (Erickson et al., 2007; Lee & Werning, 2008). This has sometimes led to speculation that these organisms could reach still larger sizes unsampled by the fossil record, but an alternate needs to be considered that this pattern is a result of paedomorphosis or peramorphic hypermorphosis – i.e., delaying sutural closure and prolonging features allowing rapid growth well into adulthood, with growth slowing but not ceasing upon sexual maturity (Lee & Werning, 2008) and cessation only occurring with senescence – which has been proposed as the mechanism by which these animals achieved such spectacular sizes in the first place (Lee & Werning, 2008; Lomax et al., 2018). A good example of this are Pliosauridae, which rarely exhibit neurocentral fusion even as adults (Araújo & Smith, 2023; Knutsen et al., 2012; McHenry, 2009) – a feature otherwise commonly used as an indicator of osteological maturity in reptiles. This suggests that some of these supposedly somatically immature individuals could be close to typical adult size and that traditional markers of somatic maturity may be less informative for megafauna (Hone et al., 2016).

Where shape estimates come from a small number of remains (e.g., Helicoprion, O. megalodon and Perucetus), life stage and sex cannot be included as confounding variables in regression and volumetric models. It is practically impossible to know the full size range of an extinct species (Mallon & Hone, 2024; Sanchez‐Villagra, 2010), and we rarely have an adequate understanding of sexual dimorphism in these taxa. Thus, the few fossil samples that do exist are simply treated as standard for their species or population. Even in the rare cases that isolated fossil remains can be distinguished as either adult or juvenile, phenomena such as pathologic gigantism (e.g., Carboniferous cephalopods, Manger et al., 1999) and insular dwarfism (e.g., Palaeoloxodon and Europasaurus, Herridge & Lister, 2012; Sander et al., 2006) make this assumption questionable. Similarly, it must be questioned whether treating the maximum sizes reached by presumably exceptional individuals as representative of the species in palaeoecological studies is as informative as using the more modest average adult size (Mallon & Hone, 2024: p. 8). For example, Mallon and Hone (2024) speculated that a hypothetical 15‐m and 15‐t ‘world record’ Tyrannosaurus rex would be so slow and require so much food that it would have to rely on scavenging or shift prey focus to sympatric titanosaurs – neither of which would be realistic behaviour for the species. Issues of ontogeny and intraspecific variation will continue to be an issue regardless of the proxy taxon used but should accounted for when considering which individuals of this proxy are to be used for size and form reconstructions, particularly where volumetric approaches are used. Ideally, studies should conduct routine sensitivity tests considering multiple models that account for variation in both ontogenetic stage and sex. Most importantly, researchers should not assume that size or form reconstructions made from a handful of incomplete remains are representative of the full range of morphology seen in an extant species, or that these remains represent the ‘average’ morphology.

4.4. Phylogenetic placement

When selecting proxies for estimating body size in extinct taxa, consideration of the phylogenetic placement of both proxy and extinct focal taxon is vital. Where proxies are selected on the basis of phylogenetic similarity, it is imperative that the phylogenetic placement of both taxa is well resolved, which is frequently not the case. Issues of phylogenetic uncertainty are systemic in palaeobiology (Marjanović & Laurin, 2007; Reeder et al., 2015). This is exemplified by O. megalodon, where C. carcharias has generally been considered the best modern proxy despite the uncertain placement of O. megalodon within lamniform sharks (Sternes et al., 2023, 2024). At best, this phylogenetic uncertainty raises doubts about the validity of specific proxy taxa and may distort the results of phylogenetic comparative analyses which are themselves used to estimate body size in some studies (Diniz‐Filho & Nabout, 2009; Paiva et al., 2022; Symonds & Elgar, 2013). In extreme cases, phylogenetic uncertainty could cloud our understanding of homology between the anatomical units that are used to predict body size and form, making it impossible to clarify the validity of proposed proxies. Of course, proxies are not always selected on the basis of phylogeny. Studies may alternatively seek to use taxa assumed to be convergent in ecological habits or body form (Engelman, 2023b; Ferrón et al., 2017). However, this too is problematic given that there is no guarantee that phylogenetically disparate groups should display similar scaling relationships between anatomical features, regardless of the perceived degree of morphological convergence. For this reason, we favour the consideration of phylogeny when selecting appropriate proxy taxa but stress that for such an approach to be valid requires well‐resolved phylogenetic placement of both the proxy and the study taxon. Ultimately, estimates of body size and form made in this way must always be treated with caution given that perceived phylogenetic relationships between these taxa are intrinsically hypothetical in nature and subject to revision upon the inclusion of new data.

4.5. Social pressures

While most spurious size/form estimates are likely driven by some combination of the factors outlined above (and a general unawareness of biostatistical best practices), social pressures and the nature of research academia also have the potential to influence reconstructions of body size/form in extinct animals. Studies reporting spectacular sizes for organisms are often widely read and publicised, which can significantly elevate the work of early‐career researchers and translate to significant opportunities for funding and public interest. Some extinct species have gained considerable media attention as a direct result of their unusual size relative to modern animals (Ferreira et al., 2024; Head et al., 2009, 2013; Molnar, 2004; Rinderknecht & Blanco, 2008; Wroe et al., 2004) and might otherwise have failed to appear in high impact journals or receive widespread public attention if they were smaller. At the same time, while journals are often eager to publish on studies suggesting spectacular sizes, more modest, revised estimates are less likely to be considered publishable as they are unlikely to garner broader interest. There also seems to be a tendency of human nature to overestimate the size of megafauna unless quantitatively measured. This is well‐demonstrated by several studies on extant megafauna noting that even experienced field biologists tend to overestimate the size of their subjects (Greer, 1974; Molnar, 2004; Randall, 1973; Wood, 1976; Woodward et al., 1995). Other palaeontologists have made similar observations. As noted by Grillo and Delcourt (2017: p. 83) in their study of abelisaurid theropods, ‘the fact that most published BL [body length; = total lengths] are overestimates reinforces a statement made by Therrien and Henderson (2007) that the lack of complete skeletal remains in large theropods gives free course to imagination, that allow researchers to present new specimens as ‘the largest’, ‘the heaviest’, or other kind of similar adjectives’. Similarly, Fuchs et al. (2020: p. 42) noted previous estimates of size and form in their study organism (Enchoteuthis) seemed to be based on the ‘hope’ (wording theirs) of a more spectacular animal rather than any fossil evidence. While it is unlikely that these factors are acting in all or even most cases of controversial size/form estimates, the current landscape of academia does potentially encourage overly generous size estimates. Consequently, in our attempt to cover all potential influences on body size/form estimates, we would be remiss to not mention these social pressures as a potential bias.

Several factors may also make researchers reluctant to publish modest size estimates of extinct taxa. Researchers may be reluctant to downsize spectacular charismatic megafauna for fear that it will reduce public interest in their research area or burn bridges in the academic community, which could have downstream consequences for collaborations, funding acquisition or even the outcome of peer review. They may also fear museums may restrict access to specimens or otherwise respond poorly to research downsizing their flagship taxon. Furthermore, one must be aware of backlash from the ever‐growing fan communities of prehistoric organisms (such as Dunkleosteus, O. megalodon and theropod dinosaurs) on the internet, who may feel strongly about the perceived appearance of their favourite organisms. None of these concerns are hypotheticals, and all have happened at one point or another to many palaeobiologists who study well‐known, iconic fossil taxa, including some of those mentioned in the present study. All studies should be judged on their scientific merit through debate and discourse, through which progressive improvements to our understanding of extinct animals can be gained.