Interelemental osteohistological variation in Massospondylus carinatus and its implications for locomotion

Introduction

Massospondylus carinatus is the most abundant non-avian dinosaur known from southern Africa, and hundreds of specimens have been referred to this taxon since its description in 1854 (Kitching, 1979; Kitching & Raath, 1984; Gow, Kitching & Raath, 1990; Sues et al., 2004). Massospondylus carinatus has been found in the upper Elliot and Clarens formations of the Stormberg Group in South Africa and Lesotho as well as in the corresponding Forest Sandstone and Mpandi Formations in Zimbabwe (Cooper, 1981; Kitching & Raath, 1984; Munyikwa, 1997; Catuneanu, Hancox & Rubidge, 1998; Bordy & Catuneanu, 2002; Rogers et al., 2004; Barrett et al., 2019). These units are all hypothesised to be Early Jurassic in age (Blackburn et al., 2013; Bordy & Eriksson, 2015; McPhee et al., 2017; Bordy et al., 2020; Viglietti et al., 2020). Massospondylus carinatus is part of the geographically widespread clade Massospondylidae, represented by medium-bodied species from four continents (Apaldetti, Pol & Yates, 2013; Chapelle & Choiniere, 2018). The large number of specimens, ranging in size from embryos to adults, makes Massospondylus carinatus an ideal taxon for studying ontogenetic variation in dinosaurs (Gow, 1990; Chapelle & Choiniere, 2018; Neenan et al., 2018; Chapelle et al., 2019a; Chapelle et al., 2019b; Chapelle, Fernandez & Choiniere, 2020; Chapelle, Botha & Choiniere, 2021). It has been hypothesised that Massospondylus carinatus underwent a locomotory shift, hatching as a quadruped and adopting bipedalism later in its development (Reisz et al., 2005). However, recent studies on vestibular system morphology, and on forelimb-to-hindlimb robusticity ratios have falsified this ontogenetic shift (Neenan et al., 2018; Chapelle et al., 2019b). Its position within the sauropodomorph phylogenetic lineage, combined with its abundance and stratigraphic age, make Massospondylus carinatus an important taxon for better understanding early dinosaur evolution and the palaeoecology of Early Jurassic ecosystems.

For nearly two centuries, osteohistology has been recognised as a useful tool for studying life history traits in dinosaurs (Lee & Werning, 2008; de Buffrénil et al., 2021; Griffin et al., 2021), including growth rates (Erickson, Rogers & Yerby, 2001; Erickson, 2005; Lehman & Woodward, 2008; Erickson, 2014; Bailleul, O’Connor & Schweitzer, 2019), metabolic rates (Sander & Klein, 2005; Erickson et al., 2009), and the onset of sexual maturity (Erickson et al., 2007; Lee & Werning, 2008). Chinsamy’s (1993) study on a size series of Massospondylus carinatus femora was the first detailed work on basal sauropodomorph osteohistology (Chinsamy, 1993; Klein & Sander, 2007). This study found that growth rate decreased with age, but that growth did not cease completely (Chinsamy, 1993; Klein & Sander, 2007). Chinsamy (1993) also concluded that the taxon displayed a cyclical, indeterminate growth strategy, based on the presence of growth marks in the cortical bone and the absence of a true outer circumferential layer (= external fundamental system, EFS), respectively. They suggested this evidenced an intermediate physiology between ectothermy and endothermy (Chinsamy, 1993; Klein & Sander, 2007).

Since then, osteohistology has also been used to examine locomotory shifts in dinosaurs by relating forelimb-to-hindlimb ratios with age (such as a shift from quadrupedalism to bipedalism in Psittacosaurus: (Zhao et al., 2013)). Osteohistological studies of intraspecific ontogenetic series have been carried out across the dinosaur tree, using different postcranial elements including: a size series of a single element such as femora (Chinsamy, 1993; Klein & Sander, 2007; Skutschas et al., 2021) or tibiae (Woodward et al., 2015); strictly or mostly stylopodial elements (Sander, 1999; Klein & Sander, 2008); and mixed postcranial elements, including limb bones and ribs (Erickson et al., 2006; Werning, 2012; Stein, Hayashi & Sander, 2013; Cerda, Pol & Chinsamy, 2014). Some studies also investigated cranial osteohistological growth, notably in ceratopsian frills (Horner & Lamm, 2011; Fostowicz-Frelik & Slowiak, 2018).

Recent work has highlighted some caveats in estimating dinosaur growth curves. Taxon misidentification, errors in age estimates and when retrocalculating missing cyclical growth marks (CGMs) or lines of arrested growth (LAGs), the lack of data points in an ontogenetic sample, as well as the scarcity of fully mature specimens all warrant caution in model-based studies of dinosaur growth (Myhrvold, 2013). Growth strategies have also been found to be divergent in dinosaurs, with both acceleration and hypermorphosis models having been hypothesised in theropods, for example (Cullen et al., 2020). Furthermore, the relationship between LAG spacing and organismal growth has been found to be inconsistent, with intraelemental variation leading to markedly different conclusions regarding the relative maturity of specimens depending on which area of the bone was analysed (Cullen et al., 2021). Further uncertainties arise when comparing different elements from a single individual due to intraskeletal variation in LAG counts (Cullen et al., 2021) as well as varying levels of secondary bone remodelling with larger bones showing less remodelling than smaller ones (Padian, Werning & Horner, 2016). Archosaurs have been found to show interelemental variation in growth rates, with the femora, tibiae and humeri growing faster than other limb elements. LAG counts also vary between elements (Curry, 1999; Woodward, Horner & Farlow, 2014). Finally, osteohistological growth plasticity (hereafter ‘growth plasticity’) further exacerbates the difficulty of estimating ages and determining growth patterns in dinosaurs. A recently published series of recommendations attempts to resolve the issues that can confound osteohistological analysis, including: increasing sample sizes for the individuals in a single ontogenetic series; incorporating tissue organization and vascularity changes into these analyses; and multielement sampling (Cullen et al., 2021).

A recent study on Massospondylus carinatus found high variability in LAG spacing that provided evidence for growth plasticity with disparate body masses for Massospondylus carinatus individuals at given skeletal ages (Chapelle, Botha & Choiniere, 2021). The documented weak correlation between age and size, coupled with the lack of extremely young and senescent individuals, warrants caution when inferring growth curves (Erickson, Rogers & Yerby, 2001; Gee, Haridy & Reisz, 2020). Similar growth plasticity has been recorded in Plateosaurus trossingensis and Mussaurus patagonicus, with histological features correlating poorly with body size and possibly being influenced by environmental factors instead (Sander & Klein, 2005; Klein & Sander, 2007; Cerda et al., 2022). Although the osteohistological growth of Massospondylus carinatus femora has been well researched, studies on the growth patterns of other skeletal elements, as well as documentation of interelemental variation, are still lacking.

A revised set of diagnostic criteria for Massospondylus carinatus (Chapelle & Choiniere, 2018; Barrett et al., 2019), new specimens referrable to the genus, and recent stratigraphic revisions (Viglietti et al., 2020) invite a reconsideration of Massospondylus carinatus growth. Here, we present the results of a broad, multielement, multiontogenetic stage study that aims to: (1) answer questions about interelemental variation during ontogeny; (2) comment on previously proposed growth curves; and (3) determine if a locomotory shift is recorded in the bone microstructure.

Methods

A destructive sampling permit (permit number 2643) was acquired from the South African Heritage Resources Agency (SAHRA) in order to sample long bones from 27 Massospondylus carinatus specimens. The sectioned sample is composed of 48 elements: 15 humeri, two radii, two ulnae, 19 femora, 8 tibiae, and two fibulae (Table 1 and Figs. 1–10).

All osteohistological sections were produced at the National Museum, Bloemfontein following standard methods (Lamm, 2013; Botha-Brink, Soares & Martinelli, 2018). Two-to-three-centimetre-thick blocks were cut out of the long bone midshafts using a handsaw and Dremel® tool. These blocks were embedded in Struers EpoFix® resin (Struers; Cleveland, OH, USA) under vacuum and dried for at least 24 h. They were then cut (using a Struers Accutom-100®; Struers; Cleveland, OH, USA) into 1.5 mm-thick cross-sections. Thick sections were adhered to 5 mm-thick glass and plastic slides with EpoFix® resin, and ground to thicknesses of several microns using a Struers Accutom-100®. Rendering was carried out in NIS-Elements 4.5 (Nikon Corp., Tokyo, Japan) under normal (NL), polarised (PL) and cross-polarised (CPL) light, using a polarizing microscope (Nikon Eclipse Ci-POL) equipped with a digital camera (DS-Fi3), and a quarter lambda plate at 530 nm. Stitched images of complete transverse sections were assembled using NIS-Elements 4.5 (Nikon Corp., Tokyo, Japan). The smallest specimen in the study, the embryonic individual BP/1/5347a (preserved in-ovo), was studied using digital osteohistological data (for the humerus, radius, ulna, femur, tibia and fibula) obtained using synchrotron radiation X-ray micro-computed tomographic (SRμCT) scanning done at the European Synchrotron Radiation Facility (Grenoble, France) with an isotropic voxel size of 13.11 μm.

Minimum circumferences were taken at the midshaft of each fore- and hindlimb bone using tailoring tape. For elements that were already embedded and sectioned, circumferences were measured using ImageJ 1.52a (Schneider, Rasband & Eliceiri, 2012). For seven individuals that only preserved a partial femur or which did not preserve a femur at all, a femoral circumference was estimated using the regression of log₁₀ (minimum femoral circumference) to log₁₀ (minimum humeral circumference), which are highly correlated in Massospondylus carinatus with an R² value of 0.9872 and a p-value of 6.5381 * 10⁻⁷ (based on eight post-hatching specimens spanning the ontogenetic size range; Fig. S1). For two specimens that lacked a humerus, a femoral circumference was estimated using the regression of log₁₀ (minimum femoral circumference) to log₁₀ (minimum tibial circumference), which are highly correlated in Massospondylus carinatus with an R² value of 0.8471 and a p-value of 3.288 * 10⁻³ (based on seven post-hatching specimens spanning the ontogenetic size range; Fig. S2).

For ease of description, the sample was arbitrarily divided into four size classes (SCs) based on relative femoral size percentage (i.e., minimum femoral shaft circumference, which best correlates with body mass in bipedal non-avian dinosaurs according to Campione et al. (2014)) when compared to the largest specimen in the sample (BP/1/4934): 0–25% are considered to be SC1 individuals; 26–50% are SC2; 51–75% are SC3; and 76–100% are SC4 (Table 1). The proportional size of the specimens relative to the skeletally mature BP/1/4934 (Barrett et al., 2019) is referred to as % size for brevity. Traditionally, the percentage comparison relative to the largest known specimen of a species (or % size) is used for a priori sorting of elements into ontogenetic stages. However, given the extreme growth plasticity seen in Massospondylus carinatus, where size is decoupled from age, we were unable to divide the bones into discrete ontogenetic categories (Chapelle, Botha & Choiniere, 2021).

Five-to-ten times magnification images were taken in the mid-cortex of each bone. Vascular canals were then identified and their surface area measured in ImageJ 1.52a (Schneider, Rasband & Eliceiri, 2012). We measured the vascular canals only, excluding the lamellae that make up the primary osteon. Proportional vascularisation was then calculated by dividing the canal surface area by the total surface area of the field of view. The mean of these measurements was then calculated for each bone (Tables 2–3 and Table S1).

Measurements of distances between CGMs were taken in the thickest part of the cortex when possible and along a single radius extending from the middle of the sections to the subperiosteal surface (Fig. 11). Regressions of circumference, cortical thickness, number of CGMs and proportional vascularisation were analysed to test the relationships and variation between elements for the different variables in the sample (Fig. 12).

An overlapping growth series of femora was selected to retrocalculate the number of LAGs missing due to resorption and remodelling (BP/1/5347, BP/1/5253, BP/1/4266, BP/1/5241 and BP/1/4934). This was used to look at the difference in growth between the humerus and femur to assess the possibility of a locomotory shift during development (Fig. 13A). Incomplete cortical surfaces of several specimens in the sample, including the largest individual (BP/1/4934), precluded the use of LAG circumference as a body mass proxy (Cullen et al., 2021). Nevertheless, comparisons of the LAG radii in the thickest part of the cortex to their circumferences provided similar metrics for inferring growth variation (Fig. S3 and Table S2).

To illustrate the extent of growth variation in the sample, and its effects on CGM retrocalculation, a maximum and minimum retrocalculated age was determined in the overlapping growth series using the minimum and maximum CGM spacing, respectively (Figs. 13B–13C and Table S3). Finally, a growth trajectory of each specimen was plotted for the humerus, femur and tibia (Figs. 13D–13F).

Body mass (BM) estimates were made by applying the formula (log₁₀ BM = 2.754 * log₁₀ [femoral circumference] − 0.683) as developed by Campione et al. (2014) for inferring BM using minimum femoral circumferences for bipedal taxa.

All linear measurements using the osteohistological sections were made using ImageJ 1.52a (Schneider, Rasband & Eliceiri, 2012), the statistical analyses and figures were generated in Microsoft® Excel® for Microsoft 365 MSO (16.0.14326.20936) 64-bit (Corporation, 2018), R Studio Version 1.3.1056 (RStudio Team, 2020) and Inkscape Version 1.0 (Inkscape Project, 2020).

Results

Discussion

All of the individuals sampled possess limb bones with fast-growing cyclical bone, as previously reported based on smaller datasets (Chinsamy, 1993; Cerda et al., 2017). Smaller individuals exhibit a highly vascularised WPC, with such high proportions of WB that the bone tissues can be referred to the FLC subcategory of WPC (Prondvai et al., 2014; de Buffrénil et al., 2021). The next smallest post-hatching specimen in the sample, BP/1/5253, does not show any signs of slowed or arrested growth, and in this individual the cortex is probably recording bone deposited within the first year of growth only.

The second-smallest post-hatchling specimen, BP/1/4376, only includes zeugopodial bones. These exhibit an annulus in the mid-cortex that indicates a temporary decrease in growth rate. In SC3 individuals, the bone tissue is still primarily WPC apart from one tibia (BP/1/4928) that contains PFB in the outer cortex. SC4 bones show a decrease in vascularisation towards the outer cortex and a transition from WPC to PFB tissue.

All bones except the two smallest specimens in the sample have CGMs, in the form of LAGs or annuli (of PFB or LB or both). None of the stylopodial bones record an EFS, but one larger SC3 specimen (BP/1/4928) has an EFS in its tibia and fibula. The femur of this specimen lacks an EFS but does exhibit a strong decrease in vascularisation in the outer cortex. This, coupled with the annuli present in the radius and tibia of the second smallest specimen, indicates that the zeugopodia record a decrease in growth before the stylopodial bones do, likely because the former grew more slowly. Several non-mutually exclusive hypotheses explain this pattern. Among them is the observation that femoral and humeral ossification lags behind that of the tibia/fibula and radius/ulna in archosaur embryos (Rieppel, 1993; Fröbisch, 2008), suggesting that embryonic ossification patterns might dictate later growth trajectories. Alternatively, macroevolutionary changes in limb proportions tend to emphasize changes in the stylopod and the autopod, whereas the trajectory of the zeugopod is more conserved (Young, 2013). Although none of the stylopodial bones have an EFS, the largest Massospondylus carinatus specimen (BP/1/4934) does show signs of slowed growth (such as several closely spaced LAGs in the outer cortex and high levels of secondary remodelling) and being near maximum size (Barrett et al., 2019).

Contrary to previous hypotheses (Chinsamy, 1993), the presence of an EFS in the slower growing zeugopodia indicates that Massospondylus carinatus had determinate growth. Instead, our results indicate that there are no completely skeletally mature specimens in the sample, consistent with the observation that all vertebrates likely have determinate growth (Woodward, Horner & Farlow, 2011). In our sample, SC1 and SC4 have the smallest sample sizes. This is consistent with the conclusions of Myhrvold (2013), who noted that very immature and very mature individuals are underrepresented in a wide variety of dinosaur clades. This could be due to preservation bias (Brown et al., 2013; Brown et al., 2021), errors in age estimation methods, or to high mortality rates in mature individuals (due to competition, disease, predation) (Erickson, 2005; Erickson et al., 2006; Hone & Rauhut, 2010).

The growth trajectory of CGM number vs CGM radius is derived from a series of femora of overlapping circumferences that enabled retrocalculation of the number of LAGs resorbed during remodelling. Based on our femoral growth series, Massospondylus carinatus would not have ceased growth before 20 years of age, although it would have been close to its maximum size from 18 years of age (according to BP/1/4934; Barrett et al., 2019). This growth trajectory (Fig. 13) is a poor fit for the sigmoidal curve hypothesised for most vertebrates (Erickson, Rogers & Yerby, 2001; Erickson, 2005; Erickson, 2014). However, the lack of an EFS in any of the femora in the sample affects the shape of this growth curve by removing a potential plateau towards the end of the ontogenetic trajectory.

Given the degree of growth plasticity already reported in Massospondylus carinatus (Chapelle, Botha & Choiniere, 2021), in which body size is hypothesised to be decoupled from age, it is difficult to assess how informative our femoral series is. There are poor correlations between element circumference (a proxy for body size) and the numbers of CGMs in several elements (Figs. 12A–12C). We demonstrate that smaller specimens can have a higher number of CGMs than larger individuals, as well as smaller distances between LAGs. Some individuals show similar CGM numbers and spacing across all limb elements (e.g., BP/1/5238 and BP/1/5241), whereas others show a lack of synchroneity (e.g., NMQR3964 and BP/1/4928) (Fig. 11). There are also poor correlations between the degree of vascularisation and element circumference (Figs. 12G–12I). This supports the earlier observation that Massospondylus carinatus exhibits growth plasticity not only in the femur (Chapelle, Botha & Choiniere, 2021) but in all its limb bones (Figs. 11 and 12D–12F). Including very young and senescent individuals would improve the strength of the correlation. However, our sample documents the most active and important parts of growth history, which show few predictable patterns.

The estimated age of 20 years old for the largest known individuals of Massospondylus carinatus is slightly less than that reported for other early branching sauropodomorphs. Plateosaurus trossingensis has an estimated maximum age of 27 years old (Sander & Klein, 2005), whereas the oldest Mussaurus patagonicus specimen has a proposed age of 29–34 years (Cerda et al., 2022). This agrees with the linear relationship between body mass and life span of poikilothermic animals, with Plateosaurus trossingensis and Mussaurus patagonicus having a larger body mass than that of Massospondylus and therefore expected to have a longer life span (Atanasov, 2005; McPhee et al., 2018). CGM spacing is so variable in Massospondylus carinatus that if the maximum distance is used, the minimum age of the largest individuals decreases to 8 years, whereas if the minimum spacing is used, the maximum age increases to 81 (Figs. 13A–13C). Although these estimates are not biologically sound, as juvenile specimens will grow at different rates than adults, they assist in demonstrating the amount of variability in CGM spacing. This reiterates the difficulty in correctly determining a growth trajectory for Massospondylus carinatus. This type of growth plasticity has been noted in two other early branching sauropodomorphs: Plateosaurus trossingensis (Sander & Klein, 2005; Klein & Sander, 2007) and Mussaurus patagonicus (Cerda et al., 2022). Both of these taxa show a decoupling between size and estimated age. In addition, Plateosaurus trossingensis is hypothesised to reach maximum size (i.e., marked by the presence of an EFS) at different ages. However, the sample includes a mix of stylopodial and zeugopodial bones, which could be obscuring growth patterns due to differences in growth rates between these elements. Mussaurus patagonicus shows that some similarly-sized individuals exhibited cyclical growth whereas others showed continuous growth. This is not the case in Massospondylus carinatus, however, as all sampled specimens in SC2 and above (those above 31.95% of maximum body size) show cyclical growth.

While variations in growth rates within a species can be due to both environmental effects and intrinsic factors, the latter seem to be between populations as a whole and relate to average growth rates per day, rather than individual seasonal growth rate variation (Ferron & Ouellet, 1991; Arendt, 1997). While intrinsic factors could explain different adult body sizes, as well as different maximum and average growth rates, we would still expect individuals of a species to show similar growth patterns and trends. The variation in yearly growth rate in all individuals seems more likely to be due to environmental factors.

Reproductive maturity is usually osteohistologically identified by the transition to slowed growth, either by the decreased spacing of growth marks, or a change in overall tissue type. In our study, there is no evidence for a steep decrease in the growth trajectory of the larger individuals (Figs. 13D–13F). Specimens BP/1/3964, BP/1/5193, BP/1/5005, BP/1/4693, BP/1/5241, BP/1/4998, BP/1/5108, BP/1/4928, BP/1/6125, BP/1/5000, BP/1/5011 and BP/1/4934 generally exhibit decreased spacing in LAGs towards the bone periphery, suggesting that these individuals might have been reproductively mature (Fig. 11). Decreased LAG spacing is more common in the largest specimens (from BP/1/4928 in the above-mentioned list, onward). The tibia of BP/1/5108, the tibia and fibula of BP/1/4928, the humerus of BP/1/6125, radius of BP/1/5011 and humerus and femur of BP/1/4934 were considered adults in this study (based on the degree of secondary remodelling, LAG spacing and number, and the amount of PFB), suggesting that these larger individuals may have been reproductively mature. It is, however, difficult to ascertain an age for the onset of reproductive maturity with confidence.

There are no distinct differences in the growth patterns, amount of vascularisation, or CGM distances between the hindlimb and forelimb of Massospondylus carinatus (Figs. 12–13). Multi-element histological studies on Psittacosaurus lujiatunensis and Mussaurus patagonicus found that the locomotory shift from quadrupedal to bipedal was supported by evidence of faster growth in the humerus than in the femur at early stages of ontogeny (Zhao et al., 2013; Cerda et al., 2022). This was also supported by evidence from independent analyses of limb proportional lengths and centre of mass modelling. In Massospondylus carinatus, the fore- and hind limb bones have similar bone tissue textures (i.e., WPC). The amounts of vascularisation in the humerus and femur of BP/1/5253, the smallest post-hatching individual, are similar (27.11% and 24.90% vascularisation, respectively). This is also the case in the radius and tibia of BP/1/4376, the second smallest individual (21.5% and 19.32% vascularisation, respectively). The vascular arrangement is similar in all four bones (mainly laminar vascularisation with some patches of longitudinal, plexiform and reticular canals). Similarly, in the more mature SC3 individual BP/1/4998, the vascularisation does not differ in any significant way between the humerus, femur, tibia or fibula (20.55%, 22.31%, 15.41% and 25.16% vascularisation, respectively). The regression of humeral vs femoral proportional vascularisation has a slope of 0.5, an adjusted R² value of 0.79 and a p-value of 1.153 * 10⁻² indicating that although the humeri and femora have differing amounts of vascularisation, this proportion remains constant throughout the growth history (Fig. 12K). Finally, in the overlapping growth series, regressions of CGM number (as a proxy for age) vs CGM radius (as a proxy for size) indicate that the humerus and femur followed similar growth trajectories throughout ontogeny (Fig. 13A). Both followed a strongly linear trajectory with adjusted R² values of 0.95 and 0.96, and p-values of 1.158 * 10⁻¹⁵ and 1.157 * 10⁻¹⁵ respectively. These values do not indicate any noticeable growth pattern changes between the fore- and hind limb bones.

None of the bones in the sample included medullary bone, an ephemeral type of cancellous bone tissue deposited along the endosteal surface of the medullary cavity as well as in the intertrabecular area of gravid female birds, in which it is resorbed for calcium use in egg and embryo development (Canoville, Schweitzer & Zanno, 2019; de Buffrénil et al., 2021). Since the probability of our sample containing only males is exceedingly low (7.45 * 10⁻⁹ assuming equal sex ratios), the absence of medullary bone implies either that Massospondylus carinatus did not lay down this tissue, or that none of the sampled females were gravid (either due to sexual immaturity, or to death outside the reproductive season). Medullary bone has yet to be identified with certainty in Sauropodomorpha (Prondvai, 2017; de Buffrénil et al., 2021), offering some support to the former hypothesis. For example, neither of the large ontogenetic samples available for Plateosaurus trossingensis or Mussaurus patagonicus preserve medullary bone.

Conclusion

Our osteohistological study of multiple Massospondylus long bones from multiple anatomical regions, ranging in size from embryo to adult, reveals substantial new information on the growth history of this early branching sauropodomorph. We show that: (1) erratic CGM spacing in all sampled elements strongly supports growth plasticity, with poor correlation between body size and CGM numbers; (2) growth trajectories inferred for an individual can, but do not necessarily, vary depending on which limb element is studied; and (3) there is no evidence for differential growth rates in forelimb vs hindlimb samples from the same individual. These findings, taken together with previous research on related taxa such as Plateosaurus trossigensis and Mussaurus patagonicus, suggest that interelemental variation and growth plasticity were widespread in Late Triassic and Early Jurassic sauropodomorphs, and further falsify hypothesised ontogenetic postural shifts in Massospondylus carinatus. Similar studies are necessary in taxa from different time periods and across the phylogenetic tree to clarify how widespread this was, as well as the causal factors behind it. This study adds to the growing body of broadly sampled ontogenetic osteohistological analyses in non-avian dinosaur taxa, which have highlighted previously unappreciated complexity in their growth patterns.

Supplemental Information