External and internal structure of ankylosaur (Dinosauria, Ornithischia) osteoderms and their systematic relevance

ABSTRACT Ankylosaurian systematics can be assessed using morphological, textural, and histological characters of osteoderms. Archosaur osteoderms have cortices surrounding cancellous cores. Ankylosaurs are united by an external cortex distinguishable from the core and by the presence of mineralized structural fibers. Nodosaurid osteoderms lack a well-developed basal cortex and have dense external cortical fibers. Ankylosaurid osteoderms are thinner than those of other ankylosaurs. Polacanthine osteoderms have a cancellous core, but share this feature with other derived and primitive taxa. Cortical thickness overlaps among groups, so a thick cortex is not diagnostic for polacanthines. Specialized elements diverge histologically from the primitive condition to suit their specific functions. Some shapes and external textures are diagnostic for specific taxa, such as Ankylosaurus and Glyptodontopelta. Parsimony analyses suggest osteodermal support for a monophyletic Polacanthinae (excluding Mymoorapelta) and Shamosaurinae. SUPPLEMENTAL DATA—Supplemental materials are available for this article for free at www.tandfonline.com/UJVP


INTRODUCTION
The ankylosaurs are a group of dinosaurs known, in part, for their extensive system of osteoderms, composed of individual elements that vary in size and shape and cover part of the skull and most of the body (Coombs, 1971(Coombs, , 1978. Although known for over 175 years, these animals are still poorly understood. The morphology of their osteoderms varies in shape, microanatomy, and external sculpturing. Thus, the osteoderms of ankylosaurs present a host of problems in taxonomic identification and conducting phylogenetic analyses. Over 40 years ago, Coombs (1971:313) wrote that dermal armor is perhaps the most frequently encountered ankylosaur fossil material, "and consequently a great deal of ink has been spilt describing individual plates," which for the most part were considered not diagnostic. Despite this dismissal, he did note some differences between nodosaurids and ankylosaurids as well as accurately detailing the basic layout of osteoderms across the body. He noted that osteoderms were arranged in transverse rows down the length of the body. According to Coombs (1971), isolated osteoderms from ankylosaurids generally exhibit excavated basal (= internal) surfaces, whereas those of nodosaurids are relatively thick and display flat basal surfaces. Also, tall, solid, conical spines (at least twice as tall as their greatest diameters) were observed only in nodosaurids (although some spines from Hylaeosaurus armatus Mantell, 1833, andSauropelta edwardsorum Ostrom, 1970, were deemed almost indistinguishable from stegosaur caudal spikes). Coombs (1971) also noted marked differences in the cervical half rings of ankylosaurids and those of nodosaurids. In ankylosaurids, the osteoderms of the cervical half rings are generally separated from one another, whereas those of nodosaurids abut, often fusing with one another (Coombs, 1971). Finally, over the pelvic region, fusion of the osteoderms can sometimes form a shield, which presented no taxonomic importance to Coombs (1971). Others have disagreed with this assessment of the pelvic shield, arguing that it is in fact a synapo-* Corresponding author. morphy for the Ankylosauria (Blows, 2001;Carpenter, 2001), a possibility supported by a recent review .
More recent studies have investigated the utility of external textures and other osteodermal features in more rigorous ways (Ford, 2000;Penkalski, 2001;Scheyer and Sander, 2004;Burns, 2008;Hayashi et al., 2010). Because osteoderms are the most common osteological elements from these dinosaurs, they have great potential for clarifying issues regarding ankylosaur taxonomy, ontogeny, and behavior. However, this can only occur by establishing a baseline for comparison. Once established, issues regarding normal morphological variation can be understood and deviant morphologies recognized.
To date, there has been little published on the histology of ankylosaur osteoderms, although several recent studies increase what we know about the internal structure of these elements. Vickaryous et al. (2001) studied Euoplocephalus, focusing on the histological interactions between the dermatocranium and overlying cranial osteoderms. De Ricqlès et al. (2001) analyzed postcranial ossicles from Antarctopelta. These authors were the first to recognize a regular organization of structural fibers in the ossicles. In addition, de Ricqlès et al. (2001) histologically discussed osteoderm skeletogenesis, suggesting that a differentiated histology indicated neoplasia rather than metaplasia (the latter would have resulted in a uniform distribution of structural fibers matching the parent dermis, which they did not observe). Barrett et al. (2002) briefly described the histology of osteoderms belonging to Polacanthus foxii and Scelidosaurus harrisonii Owen, 1861. Scheyer andSander (2004) were the first to systematically investigate variation in the histology of ankylosaur osteoderms. They showed that the tissue type and arrangement of internal structural fiber bundles differed among three groups of ankylosaurs (ankylosaurids, nodosaurids, and polacanthids). Main et al. (2005) also examined ankylosaur osteoderms and included an analysis of basal thyreophorans, tracing the evolution of ankylosaur osteoderms as modified basal thyreophoran osteoderms. Most recently, Hayashi et al. (2010) investigated specialized osteoderms (tail club osteoderms and nodosaur spines). They report that these specialized osteoderms have the same histology as those of unmodified dorsal osteoderms, and that three morphogroups can be distinguished.
There is evidence to suggest that osteodermal (and integumentary) characters are crucial to our understanding of vertebrate evolutionary relationships. A study on the phylogeny of the Amniota by Hill (2005) demonstrates the effects of increasing taxonomic and character sampling and highlights the worth of the integument as a source of meaningful morphological character data. This study is important in that the incorporation of data from the integument resolves relationships that traditionally studied anatomical characters do not, revealing phylogenetic signals that were previously obscured by incomplete taxonomic or character sampling. Carpenter (1990), Ford (2000, and Penkalski (2001) noted that osteoderm form and sculpturing have been used taxonomically in other groups such as aetosaurs, crocodylians, glyptodonts, and lizards.

Sources of Osteodermal Variation
Possible sources of variation in osteoderm histology and morphology include taxonomy, individual variation, sexual dimorphism, ontogeny, and pathology. Individual variation likely accounts for some of the diversity observable in ankylosaur osteoderms; however, there are not many ankylosaur specimens preserving in situ osteoderms. Therefore, it is difficult to test the effects of individual variation on osteoderm morphology and histology. It is safe to assume that, given emerging patterns across taxa in osteoderm morphology (Carpenter, 1990;Ford, 2000;Burns, 2008) and histology (Scheyer and Sander, 2004;Burns, 2008;Hayashi et al., 2010), real taxonomic differences are being observed.
The hypothesis of sexual dimorphism in ankylosaur osteoderms is difficult to test, given the small sample sizes available. This would also presuppose that at least some of the osteoderms have a sex-linked intraspecific function, which may be true for specialized osteoderms but seems unlikely for the most of the body osteoderms. There is evidence to suggest that this may be a source of variability in tail club osteoderms (Arbour, 2009). Carpenter (1990) has suggested a sexual selection role for anteriorly projecting distal cervical spines in Edmontonia rugosidens. This interpretation of dimorphism, however, is based on three specimens, and the apparent dimorphism is just as likely caused by individual, ontogenetic, or geographic differences. Some recent studies have investigated ontogenetic changes in osteoderms in various extant organisms such as armadillos and crocodilians (Vickaryous andHall, 2006, 2008). Both of these groups display a delayed onset of osteoderm development, and this development occurs asynchronously across different regions of the body. The same may have occurred in ankylosaurs as well, as suggested by juvenile ankylosaur material from Asia. Juvenile Pinacosaurus grangeri Gilmore, 1933, specimens develop ossified cervical half rings early in ontogeny, whereas the remaining osteoderms do not fully develop until later .

Objectives
This study describes the internal and external morphology of osteoderms from various ankylosaur taxa, emphasizing specimens associated with other diagnostic skeletal material. Previous studies (Scheyer and Sander, 2004;Hayashi et al., 2010) have relied on material lacking definitive taxonomic identification, although these papers did not investigate variation below the family level. In some cases, access to specimens for destructive analysis is limited. Characters identified via this process are then incorporated into existing character-taxon matrices to test the effects of incorporating this new data on existing phylogenetic hypotheses for the Ankylosauria as demonstrated by Hill (2005).

Terminology
Although discourse on terminology may seem redundant, a lack of consistent descriptions is one of the greatest impediments to our understanding of osteoderm variation and comparison for ankylosaurs. Although several attempts (e.g., Ford, 2000;Blows, 2001;Scheyer and Sander, 2004) have been made to do this for ankylosaur osteoderms, they have not been adopted by other workers. These studies are reviewed and revised to create a system for the consistent holistic description of osteoderm position, morphology, surface texture, and histology. Even if not adopting the terminology presented herein (Fig. 1), future studies should clearly describe and illustrate the meaning of their given terminologies to avoid the ineffectuality inherent in many descriptions.
The term 'osteoderm' unambiguously refers to a bony structure of the dermal skeleton that develops in the dermis. Other ambiguous synonyms, including 'armor,' 'dermal ossification,' 'plate,' 'osteoscute,' and 'scute' , should be avoided due to inconsistent use. The term 'median' when used in the context of a median keel or apex refers to position relative to the osteoderm itself and not to the overall anatomical position on the animal. Surface textures for different osteoderm morphologies have been detailed for only a few ankylosaur taxa (Burns, 2008). Although this feature is often included in the description of osteoderms, there is no standard. Here, terminology for describing bony surface textures follows Hieronymus et al. (2009).
'Dorsal' and 'ventral' refer to anatomical directions relative to the body (i.e., a ventral osteoderm is located on the belly of the animal). This is in contrast to similar directions specific to osteoderms, defined by Scheyer and Sander (2004) and Cerda and Desojo (2011), in which 'basal' (= 'internal';'deep' of Hill [2006]; 'proximal' of Main et al. [2005]) refers to the direction towards the deeper layers of the dermis and 'external' (= 'apical'; 'superficial' of Hill [2006]; 'distal' of Main et al. [2005]) to the direction towards the more external epidermis. These terms are confused in osteoderm descriptions because workers often fail to imagine the elements as distributed throughout the animal's dermis (e.g., 'external' is not always the same as 'dorsal').
Histologically, osteoderms are described as having a 'cortex' or two 'cortices' (external and/or basal) and a core (the terms 'medulla' and 'medullary' are not used to avoid implying that the region is associated with or similar to the medullary regions of long bones). The most frequently encountered morphology in ankylosaurs (also present in the basal thyreophorans Scelidosaurus and Scutellosaurus lawleri Colbert, 1981) includes circular or oval osteoderms ('unmodified'), each with some development of a roughly longitudinal keel or apex. Osteoderms that deviate perceptibly from such a morphology (ankylosaurid tail clubs and nodosaur cervical spines) are referred to as 'specialized.' This is an important distinction because, as will be demonstrated, specialized osteoderms can diverge from the histology/morphology otherwise characteristic for unmodified osteoderms of a group. The definition for 'ossicles' adopted here is modified from Blows (2001): small (<70 mm), amorphous mineralized dermal elements often found interstitial to major osteodermal elements.
Finally, published descriptions of ankylosaur osteoderm histology have tended to distinguish between 'Sharpey's fibers' for collagen fibers that cross the boundary of mineralization in an osteoderm, and 'structural fibers' for collagen fibers that are fully incorporated into the mineralized matrix (Scheyer and Sander, 2004;Hayashi et al., 2010). This convention is followed here, and the term interwoven structural fiber bundles (ISFB; sensu Scheyer and Sander, 2004;Scheyer and Sánchez-Villagra, 2007) is used to refer to metaplastic bone dominated by structural fibers. Other Abbreviations-CI, consistency index; ISFB, interwoven structural fiber bundles; MPTs, equally most parsimonious trees; PPL, plane-polarized light; RI, retention index; TL, tree length; XPL, cross-polarized light.

MATERIALS AND METHODS
Osteoderms were examined from North American and Asian ankylosaurid, nodosaurid, and polacanthine taxa, as well as specimens of the basal thyreophoran Scelidosaurus. External variation in osteoderm shapes and textures among these specimens was studied through measurements, observations, and photographs. In addition, previously described osteoderm specimens (Scheyer and Sander, 2004;Hayashi et al., 2010) have been reviewed from FIGURE 2. Skull, mandible, and first cervical half ring of Edmontonia rugosidens (TMP 1998.98.1). A, mounted skull, mandible, and first cervical half ring in right lateral view (anterior is to the right); B, detail of first cervical half ring in dorsal view; C, detail of first cervical half ring in right lateral view. Scale bars equal 5 cm. the literature and published figures (not first-hand specimens) where noted. Due to the overwhelming number of osteoderm specimens available for study, gross morphology of these elements is not listed for each specimen. Instead, overall descriptions for individual taxa are given.
When available, representative osteoderms were selected for paleohistological analysis. These samples were stabilized via resin impregnation using Buehler EpoThin low-viscosity resin and hardener. Thin sections were prepared petrographically to a thickness of 60-80 μm and polished using CeO 2 powder. Sections were examined on a Nikon Eclipse E600POL trinocular polarizing microscope with an attached Nikon DXM 1200F digital camera. Scans of the slides were taken with a Nikon Super Coolscan 5000 ED using polarized film. Histological measurements were taken from imaged slides, or from published figures where noted, using ImageJ 1.40g.
The relative thickness of different histological layers was evaluated via a two-sample t-test assuming unequal variances in PAST 2.12 (Hammer et al., 2001), comparing each of the three ankylosaur groups against one another. Interstitial ossicles were excluded from these analyses. Osteoderms of indeterminate taxonomic assignment were originally excluded from the tests. The addition of these, however, in a subsequent test did not affect the results, so they are included here.
The effects of adding osteodermal characters to existing ankylosaur phylogenies were tested. Test 1 examined 21 ingroup and two outgroup taxa (Huayangosaurus taibaii Dong, Tang, and Zhou, 1982, and Scelidosaurus) and a total of 63 characters, 50 cranial characters from Hill et al. (2003) and 13 postcranial characters from Vickaryous et al. (2004) (see Supplementary Data). The cranial characters were chosen from Hill et al. (2003) because they included and revised some of the characters of Vickaryous et al. (2001) and incorporated data from other studies. Scorings for Scelidosaurus are from Parsons and Parsons (2009). Nodocephalosaurus kirtlandensis Sullivan, 1999, was coded by M.E.B. and Pinacosaurus spp. were coded from . Newly identified osteodermal characters were adapted to this data matrix to create the final character-taxon matrix of 91 characters.
Test 2 included Scelidosaurus as the outgroup with 16 ingroup taxa. Data were modified from Kirkland (1998) (see Supplementary Data). Character 32 in this test combines characters 32 and 33 from Kirkland (1998) to reduce overweighting acromion morphology. Character 23 included an a priori assumption in Kirkland (1998) about the secondary loss of a jugal/quadratojugal osteoderm and is recoded as 'absent' rather than 'lost.' Characters 38-44 from Kirkland (1998) were removed due to overlap with characters from this study, leaving a matrix of 38 characters without osteodermal data and 66 characters with osteodermal data.
Data were analyzed in TNT 1.1 (Goloboff et al., 2008) with the tree bisection reconnection (TBR) swapping algorithm with 1000 repetitions for each search. All characters were treated as unordered and of equal weight. A heuristic parsimony search using random addition sequences was performed. Bootstrap and jackknife searches used the same criteria, the latter with a resampling probability of 0.36. For Bremer support analysis, suboptimal trees were retained up to 50 steps and values were mapped onto the majority rule consensus trees from the heuristic searches. Tree and matrix editing were performed in Mesquite 2.75 (Maddison and Maddison, 2011).

DESCRIPTIVE RESULTS
In the following descriptions, percentages that appear as cortical and core thicknesses represent the thickness of each histological layer relative to the overall thickness of the osteoderm in a given thin section. In a compact osteoderm (those without a trabecular core), the core and cortices can be difficult to delineate. In some cases, the core is distinguishable as a region of remodeled Haversian bone. In others, however, osteoderms can be composed almost entirely of woven-fibered bone and lack any stratification into cores and cortices. Even in a case where there is a core region, its precise delimitation from an adjacent cortex can be difficult due to the retention of primary woven-fibered bone in the core. Instances in which the cortex and core were difficult to delineate have been noted in individual descriptions.

Basal Thyreophora
Scelidosaurus-Osteoderm morphology and arrangement are well known for Scelidosaurus based on several fully articulated specimens with osteoderms preserved in situ. Most of the osteoderms are relatively thin-walled (exhibit excavated bases) and are externally relatively smooth, exhibiting sparse pitting. Osteoderms in the cervical region are not fully fused into cervical half rings, but rather are in contact via interdigitating sutures. There is no fusion of individual osteoderms anywhere else on the body. Compared with derived ankylosaurs, osteoderm morphology is fairly homogenous across the body: they are conical in shape and grade towards more elongate both posteriorly and laterally along the body. Distal osteoderms are dorsoventrally compressed and triangular. The caudal region is completely encircled by osteoderms.
UOP 03/TS2, described by Scheyer and Sander (2004), is an oval-based spine. The base is concave, making the osteoderm walls equally thick throughout. Although originally described as rugose (Scheyer and Sander, 2004), the external surface is characterized by sparse pitting and is actually smooth relative to osteoderms of some derived ankylosaurs. The compact cortices are thin (both 6%) and composed of lamellar bone invested with fibers arranged at regular angles to the osteoderm surfaces (Scheyer and Sander, 2004:fig. 12). The core is thick (88%) and entirely trabecular, containing no Sharpey's fibers or osteons.

Nodosauridae
Edmontonia-Edmontonia is represented by many wellpreserved specimens. The morphology of individual osteoderms differs in the cervical half rings. Medial osteoderms of the first and second cervical half rings in Edmontonia rugosidens are square to polygonal and have keels that diverge posteriorly (Fig. 2). Those of Edmontonia longiceps have more rounded edges and their overall shape in external/dorsal view is more posterolaterally skewed than those in Edmontonia rugosidens. Distal osteoderms of the second half ring are specialized as anterolaterally directed spines. Similar spines are found in the distal position on the pectoral half ring that exhibit various degrees of bifurcation in different specimens, but they project more laterally (as opposed to anterolaterally) in Edmontonia longiceps (Carpenter, 1990). Posterior to the pectoral half ring is a pair of distal thoracic spines, one projecting anteriorly and one posteriorly. Fusion of these distal thoracic spines is individually variable. Over the dorsum of the thoracic region, osteoderms are circular with median keels. Over the pelvis, osteoderms are transversely oval with median keels. No osteoderms are known that are preserved in association with the caudal region.
TMP 1998.98.1 (Fig. 2) includes a well-preserved skull and is unequivocally identifiable as Edmontonia rugosidens (Vickaryous, 2006). Along with the first cervical half ring, which is preserved in situ posterior to the skull, much of the postcranium is preserved. All osteoderms possess the same surface texture-strong, uniform, pitted rugosity with sparse, reticular neurovascular grooves, and normal to obliquely oriented neurovascular foramina.
Several ossicles and one osteoderm (a flat oval) were selected for thin sectioning from this specimen. The osteoderm ( Fig. 3A-J) is composed almost entirely of ISFB. In a centrally located section (Fig. 3G, H), a core (53%) is visible as a region of diffuse primary and secondary osteons and resorption cavities. Both the thick external (24%; Fig. 3E, F) and basal (23%; Fig. 3I, J) cortices are dominated by ISFB. However, there is no clear demarcation (e.g., a continuous resorption line) between the core and cortices. Therefore, delimitation of the core is more subjective in this specimen and is identified as a region of denser osteons and resorption cavities than in the cortices.
The ossicles (Fig. 3K-Q) associated with TMP 1998.98.1 show less differentiation than the larger osteoderm. They are compact structures composed entirely of ISFB, with few primary osteons scattered diffusely throughout the elements. Near the external/basal margins, zonation is distinct and overprints the ISFB pattern (Fig. 3N, O). In all of the ossicles, structural fiber bundles are orthogonally arranged throughout, unlike the radial pattern described for the ossicles of Antarctopelta (de Ricqlès et al., 2001).
Two nodosaur distal spines, identified as Edmontonia sp. (DMNH 2452 andTMP 1979.147.94), are described by Hayashi et al. (2010:figs. 4, 5). DMNH 2452 is characterized by a smooth external surface texture with a sparse reticular pattern of neurovascular grooves. The base is flat. The external cortex (14%) is composed of ISFB-dominated bone and lacks zonation. The basal cortex (2%) is poorly developed. The thick core (84%) is trabecular and invested with neurovascular canals that connect neurovascular foramina on the base of the osteoderm with the grooves on the external surface. In the external cortex, two distinct systems of dense orthogonally arranged structural fibers are observable under XPL. Some secondary osteons are visible where the external cortex contacts the core (Hayashi et al., 2010:fig. 4).
TMP 1979.147.94 was sectioned in four locations along its length by Hayashi et al. (2010:fig. 5). The external surface is ornamented with sparse, uniformly distributed pits. Only one neurovascular groove/foramen is visible near the apex. Histologically, it lacks a basal cortex and has a trabecular core. The cortex is relatively thinner near the base than at the apex (Hayashi et al., 2010:fig. 5).
Panoplosaurus-Although no osteodermal specimens of Panoplosaurus have been sectioned to date, diagnostic morphological characters may be derived from the holotype (CMN 2759). External osteoderm texture is pitted with a dense reticulate pattern of neurovascular grooves. Medial osteoderms of the cervical half rings are suboval and the posteriorly diverging keels curve laterally. Lateral and distal cervical osteoderms are transversely elongate and have relatively high, sharp keels. The arrangement of the osteoderms on the remainder of the body is not known (Carpenter, 1990). Glyptodontopelta-Three osteoderms from SMP VP-1580 were obtained from a referred specimen of Glyptodontopelta. SMP VP-1580 C is a partial keeled osteoderm (Morphotype A or B, sensu Burns, 2008), whereas SMP VP-1580 D (Fig. 4) and E are flat (Morphotype D). The texture of each of the three is characterized by a smooth surface and a dense pattern of reticular neurovascular grooves. Neurovascular foramina are oriented obliquely to the surface. These osteoderms all share the basic histology, which is similar to the larger osteoderm described for TMP 1998.98.1 (Edmontonia) in that cores are visible in some osteoderms as concentrations of osteons and resorption cavities within the central areas of the sections; however, a solid demarcation between cortex and core is not clear, and in some specimens, ISFB dominates throughout (Fig.  4G, H). The primary bone in all osteoderms is composed of ISFB ( Fig. 4E-J). Burns (2008) sectioned a fragmentary medial cervical osteoderm from SMP VP-1580 (Morphotype F). This specimen has a compact external cortex (dominated by ISFB) and trabecular core, but lacks a basal cortex.
Sauropelta-A skeleton of Sauropelta (AMNH 3036) is among the most complete nodosaur specimen known and preserves a complete series of articulated osteoderms (Carpenter, 1984). AMNH 3035 completes the cervical series and has a partial skull. Each of the three half rings have two paired osteoderms: medial and lateral. Medial osteoderms are oval with rounded posterior apices. Lateral osteoderms are specialized as transversely elongate triangular spines that project posterolaterally, each with a sharp keel. A larger lateral thoracic spine posterior to the half rings matches this morphology. Thoracic and caudal osteoderms resemble medial elements of the half rings but are more circular. Apices on these osteoderms become taller in lateral and distal elements. Pelvic osteoderms are also circular.
Two osteoderms from a specimen of Sauropelta (DMNH 18206) were sectioned by Hayashi et al. (2010: fig. 6). One is keeled, whereas the other, smaller (roughly one-quarter the size) osteoderm is circular with an offset apex. Both have uniform, pitted, rugose surfaces and sparse, reticular patterns of neurovascular grooves. The base is concave in each. Also, neither specimen has a basal cortex; the thick (86 ± 3%) trabecular core is instead exposed at the base. The external cortex is, by contrast, well developed, of average thickness (14 ± 3%), and is largely composed of azonal ISFB. Secondary osteons are found at the boundary between cortex and core; there are also several, however, near the base. Structural fibers are dense in the external cortex and arranged either perpendicular or parallel to the surface (Hayashi et al., 2010: fig. 6).
Nodosauridae Indet-TMP 1967.10.29 is an isolated, keeled nodosaurid osteoderm from the Upper Cretaceous Dinosaur Park Formation, Alberta, Canada, that was sectioned by Scheyer and Sander (2004: fig. 6). It is circular with an offset apex. The external cortex (8%) is composed of ISFB. Structural fibers in this region are dense and arranged in two sets of orthogonal meshworks set at 45 • to one another. The trabecular core is thick (92%). Scattered secondary osteons are also found more basally in the external cortex (Scheyer and Sander, 2004: fig. 6).

Ankylosauridae
Ankylosaurus-There are few specimens of Ankylosaurus (Carpenter, 2004); therefore, none were available for thin sectioning. Osteoderms of this taxon are smooth and each has a distinctive pattern of prominent but sparse neurovascular grooves. The size and depth of each groove approach the texture seen on many ceratopsian frill and craniofacial bones. Most of the osteoderms are flat, and only a few have low keels near the lateral margins. Others include median keeled and flat circular osteoderms (Carpenter, 2004:figs. 19, 20).
Nodocephalosaurus-SMP VP-2067 (Fig. 5) is a referred specimen that was found near the same locality and at the same stratigraphic horizon as other specimens referred to Nodocephalosaurus (Sullivan and Fowler, 2006). Being a fragment, the overall shape is unknown. The external surface texture has uniformly distributed, projecting rugosities, and sparse distribution of reticular neurovascular grooves. Neurovascular foramina penetrate the surface normally to obliquely. The external (13%; Fig. 5E, F) and basal (8%; Fig. 5I, J) cortices are thin and composed of ISFB. The thick cortex (79%; Fig. 5G, H) is composed of a mixture of trabecular (identified by larger resorption cavities) and Haversian (identified by accumulations of secondary osteons) bones but also includes some interstitial primary ISFB.
Much of the confusion surrounding the status of Campanian North American ankylosaurid specimens has been resolved in several recent revisions Blows, 2012, Arbour andPenkalski, in press); therefore, sampling was restricted here to specimens that could be confidently assigned to Euoplocephalus. UALVP 31 includes a skull and has been assigned to Euoplocephalus (Vickaryous and Russell, 2001;Arbour and Currie, 2013). Although its body placement is not known, the osteoderm studied (Fig. 6) was recovered in association with the pelvic region. The external surface texture is smooth with no neurovascular grooves or foramina. The cortices are of average thickness (22 ± 3%) and composed of azonal bone tissue, with dense structural fibers approaching orthogonal arrangements near the external and basal surfaces (Fig. 6E, F, I, J). The core (57%; Fig. 6G, H) grades smoothly into the cortices and consists predominantly of trabecular bone, although spicules are relatively thick and do retain some primary ISFB.
Several interstitial ossicles (Fig. 6C, D, K-N) of UALVP 31 were also sectioned. Their histology is similar to that of the ossicles described here for Edmontonia. They are compact structures composed almost exclusively of structural fibers arranged orthogonally throughout their entirety. A strong pattern of zonation (Fig. 6K, L) is visible near the margins of each ossicle.
Pinacosaurus- Scheyer and Sander (2004) sectioned an osteoderm fragment from Pinacosaurus grangeri (ZPAL MgD-II/27). Its overall morphology cannot be determined, but the surface is smooth, lacking neurovascular foramina and grooves. The external (6%) and basal (17%) cortices are compact. The core (77%) is largely composed of trabecular bone, although Haversian bone does occur at the junction of the core and the cortices. Structural fibers are extensive throughout the osteoderm (Scheyer and Sander, 2004: fig. 8C, D).

Other Ankylosaurs
Gargoyleosaurus-Two cervical half rings of Gargoyleosaurus are composed of six partially fused, paired, keeled osteoderms (an unfused median osteoderm had been reported by Kilbourne and Carpenter [2005], but is reinterpreted here as one of a pair of medial osteoderms because an odd number of osteoderms is unknown in any thyreophoran cervical half ring). Flat and keeled oval osteoderms characterize the thoracic region. Laterally, osteoderms of this region become elongate and triangular with excavated bases. Pelvic osteoderms have flat to apical circular morphologies. Larger osteoderms are interspersed by smaller osteoderms of identical morphology and are fused to form a continuous shield covering the pelvis .
One keeled osteoderm was described by Hayashi et al. (2010) and is associated with the holotype of Gargoyleosaurus parkpinorum (DMNH 27726). The external surface texture is smooth and lacks neurovascular grooves and foramina. The base is flat. The external (7%) and basal (11%) cortices are relatively thin and consist of ISFB. Few secondary osteons are scattered throughout the cortex, but are concentrated near the border with the core. The core (82%) is trabecular. Structural fibers are dense throughout the cortices and exhibit an orthogonal arrangement.
Gastonia-Two cervical half rings of Gastonia have triangular plates distally, although no complete or articulated half rings have been described. Posteriorly, Gastonia possesses the same osteoderm morphologies as seen in Gargoyleosaurus. Osteoderms in the pelvic region are similarly fused into a continuous shield (Kirkland, 1998;Arbour et al., 2011). Unique to the taxon are elongate, keeled spines that twist ∼90 • towards their apex (Kirkland, 1998).
Three specimens from DMNH 53206 were collected from a monospecific bone bed assemblage and, as such, can be confidently assigned to Gastonia (they are labeled A, B, and C for convenience). DMNH 53206 A is a lateral or distal spine, whereas DMNH 53206 B (Fig. 7A, B) and DMNH 53206 C are both circular osteoderms with central apices. Each of the three specimens is characterized by a smooth external surface texture and a lack of neurovascular grooves and foramina. DMNH 53206 A possesses a thin (3 ± 3%) compact cortex. The thick (92%) core is composed of trabecular bone. The few osteons preserved in the cortex are secondary. The other two osteoderms also have trabecular cores (60 ± 15%); however, the cortices are relatively thicker (14 ± 10%) in both DMNH 53206 B and C to the point where the cores are pinched out laterally. The cortices in each are composed of orthogonally arranged structural fibers. A single neurovascular foramen is visible in cross-section in DMNH 53206 C. It shows no bone modification surrounding it, but cross-cuts the structural fibers in the osteoderm.
DMNH 49754-1, DMNH 49754-4, IPB R481, and IPB R482 are osteoderms from the same monospecific bonebed assemblage. DMNH 49754-1 and DMNH 49754-2 were sectioned by Hayashi et al. (2010), and IPB R481 and IPB R482 were described by Scheyer and Sander (2004). DMNH 49754-1 is a circular osteoderm with an offset apex, whereas DMNH 49754-4 represents a spine. Most of the surface textures previously described for these specimens occur on the margins and basal surfaces. The external surface texture is smooth and lacks neurovascular grooves and foramina. The basal and external cortices in both are composed of compact bone with the same orthogonal structural fibers seen in the other specimens of Gastonia. Secondary osteons are rare throughout the osteoderms but, when present, occur at the transition between the cortex and core. The cortex is relatively thicker in DMNH 49754-4 (19 ± 3%) than it is in DMNH 49754-1 (9 ± 4%); however, the cortical thickness in the spine (DMNH 49754-4) varies from 22% at the base to 16-18% closer to the apex (Scheyer and Sander, 2004: fig. 3A-C).
Mymoorapelta-The hypodigm material of Mymoorapelta includes five distinct osteoderm types similar to those of Gargoyleosaurus and Gastonia: elongate spines with excavated bases, dorsoventrally compressed and basally excavated distal triangular osteoderms, a smaller solid spine, flat and keeled thoracic osteoderms, and a fused mosaic of pelvic osteoderms. MWC 211 (Fig. 7C-F) is a circular osteoderm with an off-center apex and an excavated base, from the Mygatt-Moore quarry (middle Bushy Basin Member, Morrison Formation) along with all other currently published material of the genus. The external surface texture is characterized by uniform, weak pitting and an absence of neurovascular grooves and foramina. Internally, the external (31%) and basal (19%) cortices are thick and consist of ISFB. Primary osteons are absent but several secondary osteons are scattered throughout the cortex. The core (50%) consists of trabecular bone. Structural fibers are dense and found in both the cortex and remnants of primary bone in the core. In the cortex, these fibers are arranged orthogonally both to the osteoderm surface and relative to one another.
Polacanthus-Cervical half rings in Polacanthus are composed of separate keeled osteoderms fused to underlying bands of bone. Spines of the anterior thoracic region and dorsoventrally compressed distal osteoderms are similar to those of Gastonia and Mymoorapelta. A continuous shield of fused osteoderms also exists over the pelvic region.
One osteoderm from Polacanthus foxii (NHMUK R9293) was sectioned and described by Scheyer and Sander (2004). It is circular with an off-center apex and a convex base. The external and basal cortices are equally thick (14%) and composed of ISFB. Structural fibers are denser in the basal cortex than they are in the external cortex, and approach an orthogonal arrangement at the basal margin of each osteoderm. In the external cortex, the fibers are perpendicular to the surface. The core (72%) is composed of trabecular bone, but a few secondary osteons are visible where the core contacts the cortex. Interstitial primary bone in the core retains structural fibers (Scheyer and Sander, 2004: fig.  3D-F).
The second search (Fig. 9B) for this set used 91 combined cranial, postcranial, and osteoderm characters (78 parsimony informative) and returned six MPTs (TL = 199, CI = 0.538, RI = 0.687). The addition of osteodermal characters has little effect on resolution but increases overall branch support for all major clades, including a sister-group relationship of Shamosaurus scutatus Tumanova, 1983, and Tsagantegia longicranialis Tumanova, 1993. Pinacosaurus is moderately supported as monophyletic in the first search. In the second, Edmontonia and Pinacosaurus are well supported as monophyletic genera, and the Nodosauridae is moderately supported as monophyletic. The addition of osteodermal characters creates a clade including Sauropelta and Silvisaurus condrayi Eaton, 1960. The basal position of Gargoyleosaurus is not supported by osteodermal characters, and its relationship to either Nodosauridae or Ankylosauridae is ambiguous.

Test Set 2
The first search using Test Set 2 (Fig. 9C) examined 38 characters (33 parsimony informative) and returned four MPTs (TL = 62, CI = 0.729, RI = 0.846). The second search (Fig. 9D) of 66 characters (50 parsimony informative) returned 12 MPTs (TL = 117, CI = 0.628, RI = 0.702). In this test, the addition of osteodermal characters increases resolution at the expense of overall branch support. Low taxonomic and character sampling likely plays a role and the osteodermal characters reveal affinities more readily than in Test Set 1. Without osteodermal characters, FIGURE 8. Proportional osteoderm cortical thickness for three groups in the Ankylosauria. The percent thickness (mm) of the cortex (proportion of combined thickness of the external and basal cortices versus total osteoderm thickness) is plotted against the total thickness (mm) of the osteoderm. FIGURE 9. Effects of the addition of osteodermal characters into existing hypotheses of ankylosaurian phylogeny. A, Test 1 included 63 cranial and postcranial characters (57 parsimony informative) from Hill et al. (2003) and Vickaryous et al. (2004) and returned 20 MPTs (TL = 149, CI = 0.480, RI = 0.672). B, the addition of osteodermal characters created a matrix of 91 (78 parsimony informative) and returned six MPTs (TL = 199, CI = 0.538, RI = 0.687). C, Test 2 included 38 characters (33 parsimony informative) from Kirkland (1998) and returned four MPTs (TL = 62, CI = 0.729, RI = 0.846). D, the addition of osteodermal data created a matrix of 66 characters (50 parsimony informative) returned 12 MPTs (TL = 117, CI = 0.628, RI = 0.702). Fifty percent majority rule consensus trees presented. Support values above branch are majority rule percentages and Bremer support values. Those below are bootstrap and jackknife support. Blanks indicate Bremer support below 1 and bootstrap/jackknife support below 50. Higher taxa listed to left of trees correspond to nodes indicated by black circles. the positions of Minmi, Mymoorapelta, and Shamosaurus are unresolved. Including these characters recovers Minmi as a basal ankylosaur. Shamosaurus is the most basal ankylosaurid, with Mymoorapelta nested deeper, as sister to the clade Polacanthinae + Ankylosaurinae.

Osteodermal Characters
Most unmodified ankylosaur postcranial osteoderms possess external cortices that are distinguishable from the cores, and have structural fibers in at least part of the osteoderms. Mature nodosaurid osteoderms can be distinguished from those of other ankylosaurs on the basis of poorly developed or absent basal cortices. Most of the larger osteoderms examined have trabecular cores, but a few exhibit more compact cores, consisting of a mixture of primary ISFB bone, scattered osteons, and resorption cavities. Likewise, an external cortex dominated by ISFB is common. This layer is complete with a dense pattern of orthogonal structural fibers, as demonstrated by Scheyer and Sander (2004) and Hayashi et al. (2010). Ankylosaurid osteoderms, on average, are thinner than those of other ankylosaurs. Some possess compact Haversian bone in their cores, whereas in others this layer has been remodeled into trabecular bone.
Ossicles-The histology of interstitial ossicles is variable, even within a single individual, although they do have some patterns. They are predominantly composed of primary ISFB arranged in orthogonal patterns. Although a radial pattern of structural fibers has been reported for Antarctopelta (de Ricqlès et al., 2001), such was not observed in the taxa sampled here. It is possible that ossicles represent different stages of osteoderm development. In addition, these elements show no consistent histological differences between ankylosaurids and nodosaurids. Both specimens examined (TMP 1998.98.1 and UALVP 31) show predominantly compact ISFB and strong marginal zonation. Ossicles may represent incipient osteoderms that stopped mineralizing before reaching maturity, or they may be developmentally distinct from larger body osteoderms. Either way, because they do not exhibit the taxon-specific characters seen in the larger osteoderms, they provide no taxonomic information.
Specialized versus Unmodified Osteoderms- Hayashi et al. (2010) examined specialized osteoderms. They concluded that spines and clubs maintain the same characteristic features for their respective clades despite the differences in sizes and morphologies. This study came to the opposite conclusion-specialized osteoderms have a correspondingly modified histology. This is supported by evidence from extant Caiman osteoderms, in which even relatively minor changes in shape can produce significant changes in histology even in a single individual (Burns et al., 2013). Polacanthine and nodosaur spines exhibit lower relative cortical thicknesses than unmodified osteoderms. In ankylosaurids, although the unmodified body osteoderms may be relatively thin and have cores composed of Haversian bone, those of the tail club knob are thick and have trabecular cores (Arbour, 2009;Hayashi et al., 2010: fig. 10). Therefore, osteoderms that are specialized to perform specific functions provide morphological characters, but their histology may be taxonomically uninformative.
Structural Fibers-The presence, density, and arrangement of mineralized structural fibers have been previously identified by Scheyer and Sander (2004) as characteristic of three distinct ankylosaur groups. Their conclusions are supported here. In nodosaurids, there are alternating three-dimensional layers of structural fibers in the external cortex: one with fibers arranged parallel and perpendicular to the surface, and one with fibers arranged 45 • . Ankylosaurids also possess structural fibers in the cortex that insert perpendicularly into the osteoderms, but become more diffuse in the core because they are generally remodeled to form secondary bone. Whereas fibers in the core are randomly distributed in polacanthines, they attain a more regular arrangement near the margins, although they are not as highly organized as in nodosaurids.
Cortical Relationships-The distinction of ankylosaurids and polacanthines based on the relative development (thickness) of the cortex (Scheyer and Sander, 2004) is not supported here. Statistical tests show that there is significant overlap in the overall thicknesses of the cortices. In nodosaurs, the basal cortex contributes little to the overall osteoderm thickness or is absent entirely. This is consistent with the findings of Scheyer and Sander (2004) and Hayashi et al. (2010). When present, this cortical layer is made up of ISFB. Because this represents a significant difference between nodosaurs and all other ankylosaurs, it is treated here as a synapomorphy for the group.
Core Histology-There is more overlap in core histology than previously suggested. Scheyer and Sander (2004) distinguished ankylosaurid osteoderms based on a Haversian core. Hayashi et al. (2010), on the other hand, identified trabecular cores in ankylosaurids, in one case examining the same thin section (TMP 1985.36.218/1) as Scheyer and Sander (2004). The discrepancy is here attributed to the highly vascular nature of the compact bone in ankylosaur osteoderms and their random arrangement of Haversian canals, supporting the interpretation of Scheyer and Sander (2004). A mixture of histologies is found in ankylosaurid osteoderms-whereas some are Haversian, others are composed of trabecular bone. This may be based on the fact that core histology is more dependent on osteoderm shape, function, and/or ontogeny than on taxonomic position. Nodosaurids, although their cores are largely trabecular, can have compact cortical bone present as in some of the osteoderms known from Edmontonia and Glyptodontopelta. Only the polacanthines examined exhibit a consistent trabecular core histology. Therefore, the retention of compact bone in the osteoderm core is a characteristic of derived nodosaurids and ankylosaurids. Due to the overlap among these higher taxa, however, it is not a basis on which to distinguish ankylosaur groups.
Osteoderm Thickness-Osteoderms of ankylosaurids have often been described as 'thin-walled' or having 'excavated' bases (Coombs, 1971). These terms have been alternately considered diagnostic (Coombs, 1971) or as subjective assessments, with too much overlap with other taxa to be of real value (Burns, 2008). The sample of ankylosaurid osteoderms examined here are significantly thinner than the nodosaurid osteoderms. Whereas the polacanthine osteoderms were also found to be significantly thicker than the ankylosaurid osteoderms, our analyses do not include distal polacanthine spines. Therefore, little confidence can be placed in this difference between ankylosaurids and polacanthines at present.
External Surface Texture-Burns (2008) differentiated North American ankylosaurids and nodosaurids at the species level based on the textures of the external osteoderm surfaces. The increased sample size in this study shows that there is more overlap among these textures than previously suggested, and only a few derived taxa may be distinguished based on this character alone. Polacanthine osteoderms are consistently smooth and lack (for the most part) neurovascular grooves and foramina. Nodosaurids may have smooth or uniformly pitted osteoderms and can develop reticular patterns of neurovascular grooves and foramina. As an autapomorphy, Glyptodontopelta has a dense pattern of neurovascular grooves, as opposed to the sparse patterns seen in other nodosaurids. Ankylosaurids may also retain smooth osteoderms but can develop strongly projecting surface rugosities. This distinguishing feature of Nodocephalosaurus may represent a taxonomic difference, or variation in epidermal covering. In addition, the characteristic pattern of prominent yet sparse neurovascular grooves in Ankylosaurus (Burns, 2008) is supported as an autapomorphy for the genus.

Osteoderm Development
Osteoderm growth in ankylosaurs is far from understood. In the absence of a growth series for any member of the Ankylosauria (let alone a complete series with associated osteoderms), comparative material must be relied on to hypothesize the mode(s) of osteoderm skeletogenesis. In crocodilians, osteoderms develop via the direct transformation of preexisting dense irregular connective tissue in the dermis (= tissue metaplasia; Vickaryous and Hall, 2008). Evidence of extensive mineralized inclusions from the dermis suggests that that is also the case in the osteoderms of ankylosaurs. Their high level of organization in the fully mineralized osteoderm is indicative of the preexisting organization of collagen bundles within the dermis, likely the stratum compactum, prior to metaplastic mineralization (Moss, 1972;Vickaryous and Hall, 2008;Sire et al., 2009;Cerda and Powell, 2010).
Ankylosaur osteoderms also have delayed onset of skeletogenesis relative to the remainder of the body skeleton, as is the case in Stegosaurus (Hayashi et al., 2009). Modern archosaurs also exhibit this delay (Vickaryous and Hall, 2008), their osteoderms appearing only well after hatching. Juvenile ankylosaurs (i.e., Pinacosaurus;  do not have osteoderms posterior to the cervical half rings. It is presumed, therefore, that their osteoderms similarly have delayed skeletogenesis as in other sauropsid groups.

Ankylosaur Integument
Osteoderms, prior to bone remodeling if it occurs, represent mineralization of soft tissue, and in essence are an in vivo mineralization of the dermis of an animal (Moss, 1969;Moss, 1972). Therefore, these structures present unique opportunities to study the soft-tissue histology of extinct organisms. Tetrapod osteoderms primitively share a common site of origin at the interface between the stratum superficiale and stratum compactum in the dermis ). In some cases, the osteoderm may extend into the stratum compactum itself. Given the densities of structural fibers in ankylosaur osteoderms, there is a relatively greater contribution from the stratum compactum. This is in marked contrast to the situation in osteoderm-bearing crown members of the Archosauria. Crocodilian osteoderms are localized in the stratum superficiale, and only the most basal margins may contact the stratum compactum (Martill et al., 2000;Salisbury and Frey, 2000;Vickaryous and Hall, 2008;Vickaryous and Sire, 2009). It is possible that the stratum compactum is relatively thick in ankylosaurs (and possibly other dinosaurs). Based on the largest dinosaur osteoderms, which belong to titanosaurid sauropods, the thickness of dermis is likely more than twice that of modern elephants (Haynes, 1991;Dodson et al., 1998).
Although there is some variation in surface textures, all of the specimens in this study can be best interpreted as having cornified sheathes as described by Hieronymus et al. (2009). This suggests that whereas there would have been some soft dermal/epidermal component covering the osteoderm, it was relatively minor compared with the keratinized structure overlying it. In extant correlates (e.g., bovid horns), the bony attachments of these structures are characterized by densely concentrated metaplastic dermal collagen fibers. These fibers also interact obliquely with the bone surface and may approach an orthogonally arranged network.
The histological condition in Scelidosaurus can be considered as the basal condition for the Ankylosauria because Scelidosaurus is likely the sister to all ankylosaurs (Carpenter, 2001). These osteoderms also exhibit histologies that ally them with the basal archosaurian condition with a circumferential cortex enclosing a trabecular core . In most other osteoderm-bearing archosaurian lineages (aetosaurs, crocodylomorphs, parasuchians, etc.), a conservative yet distinct pattern of external pitting is common. In the most basal thyreophorans, this is not the case. Instead, the surfaces of the osteoderms are relatively smooth. This smooth texturing is retained in polacanthines, whereas the pattern diverges in derived ankylosaurids and nodosaurids.
Histological similarities among ankylosaur osteoderms from each of the three groups examined indicate similar development and evolution, as hypothesized by Hayashi et al. (2010). However, morphologically distinct (specialized) osteoderms were elaborated by modification of the external morphology and internal histology (contra Hayashi et al., 2010). Primitive, basal thyreophorans lack extensive structural fibers in their osteoderm cortices (de Buffrénil et al., 1986;Scheyer and Sander 2004;Main et al., 2005;Hayashi et al., 2009Hayashi et al., , 2010. Hayashi et al. (2010) concluded that the presence of such systems in ankylosaurs indicates that either their osteoderms evolved differently or their skin differed from that of other thyreophorans. This study lends support to the former hypothesis because it demonstrates that the integument does not need to be different in order to produce these divergent histologies. It is likely that greater contributions from the dense connective tissues of the stratum compactum were involved in osteoderm skeletogenesis. Therefore, ankylosaurs utilized different developmental pathways to achieve the diversity observed in their osteoderms.

Phylogenetic Tests
Overall, the inclusion of osteodermal characters into phylogenetic analyses of the Ankylosauria can increase branch support and/or resolution, depending upon taxonomic sampling. In Test Set 2, inclusion of osteodermal characters retains a monophyletic Polacanthinae within Ankylosauridae; however, it reduces branch support for the clade. This may be caused by an affinity of Mymoorapelta to the Polacanthinae, as suggested by Kirkland (1998) based on osteodermal characters. An affinity of Shamosaurus to Tsagantegia is also recovered in Test Set 2. This is supported in Test Set 1 by the recovery of Shamosaurus + Tsagantegia, a clade for which osteodermal characters increased branch support. This demonstrates support for Shamosaurinae and Polacanthinae based on cranial/postcranial and osteodermal characters. The inclusion of Mymoorapelta in the Polacanthinae is also plausible, but will require further data. These tests do not suggest a nodosaurid affinity for the Polacanthinae in contrast with the results of Thompson et al. (2012), which included more taxa and recovered polacanthines as a grade of basal nodosaurids. That analysis examined more taxa and characters than those used here, but included only 16% osteodermal characters, compared with 31% and 42% (Test Sets 1 and 2, respectively) here.
Cranial characters have been utilized in systematic studies of the Ankylosauria and have proven to be informative (Hill et al., 2003;Vickaryous et al., 2004). The postcrania and osteoderms, however, have received relatively little attention. This study demonstrates that major anatomical systems besides the cranium are useful in systematic studies of ankylosaurs. A combination of cranial characters, the osteodermal characters identified herein, and a new suite of postcranial characters should be included in such a study.

CONCLUSIONS
Osteoderm morphology is systematically useful at the familial, generic, and specific levels. The lateral cervical and thoracic spines of nodosaurids are supported as valid taxonomic indicators for that clade. Absolute osteoderm thickness is a useful character for distinguishing between nodosaurids and ankylosaurids, the latter of which have thinner osteoderms. Polacanthine ankylosaurs share a mixture of primitive and derived characters, including lateral spines (shared with nodosaurids) and basally excavated, dorsoventrally compressed triangular distal osteoderms (shared with ankylosaurids).
The histology of an ankylosaur osteoderm is somewhat dependent on the specific function(s) of the element itself; however, there are several characters that are useful for higher-level taxonomy. Only unmodified body osteoderms are histologically useful as familial indicators. Interstitial ossicles are histologically homogenous across ankylosaur taxa. Specialized osteoderms that perform specific functions (e.g., major tail club knob osteoderms) do not always retain the basic histology of unmodified osteoderms. The overall thickness of the cortex (absolute or relative) is a character that overlaps considerably among ankylosaurs. The retention of compact bone in the core of the osteoderm is characteristic of derived nodosaurids and ankylosaurids; however, it is not a basis on which to distinguish ankylosaur groups due to overlap among disparate taxa. An osteoderm that retains a compact core can be positively identified only as belonging to a derived taxon. The basal cortex of a nodosaurid osteoderm is generally absent or poorly developed. Structural fiber arrangement is also a useful character. Nodosaurids have two three-dimensional sets of these fibers in the external cortex. Ankylosaurids have perpendicularly inserting fibers in the cortices of their osteoderms, but they become more diffuse in the core. Fibers can have a more regular arrangement near the margins in a polacanthine osteoderm.
The external surface textures of ankylosaur osteoderms exhibit more variability than previously thought. Nonetheless, certain textures are observable only in certain taxa, and can be considered autapomorphic for them. The primitive thyreophoran surficial condition (also retained in primitive ankylosaurs) is that of relatively smooth osteoderms, with only isolated patches or sparsely distributed rugosities, foramina, or grooves. A relatively smooth surface coupled with dense reticular neurovascular grooves is diagnostic for the nodosaurid Glyptodontopelta.
Smooth osteoderms with sparse but prominent neurovascular grooves are characteristic of Ankylosaurus. Based on bone surface textures in extant correlates, ankylosaurs likely had a relatively thick, keratinized sheath covering each osteoderm, similar to the condition in the modern horn of a bovid.
Our data support the interpretation of Scheyer and Sander (2004) that evidence of extensive mineralized inclusions from the stratum compactum indicates that direct metaplasia is responsible for osteoderm development in ankylosaurs; it had a greater role than in other taxa such as crocodilians. Given their structure, ankylosaur osteoderms are optimized towards a primary protective function. The osteoderms of crocodilians may not have a primary protective function, and their lack of mineralized structural fibers corroborates this. Despite this, examination of other extant and extinct taxa reveals that a single, panoptic function for the osteoderms of a species or group is rarely a feasible hypothesis. Therefore, this study cannot reject the possibility of other, secondary osteoderm functions for ankylosaurs. These may include thermoregulation, or intra-and interspecific display.
Except for bones of the dermatocranium, the dermal skeleton (and the integument in general) is a major anatomical system that has been historically underrepresented in the morphological systematics of vertebrates. This study supports the conclusion of Hill (2005), that this system does provide meaningful character data. Far from homogenous elements, ankylosaur osteoderms were physiologically active tissues that were morphologically and histologically optimized for specific functional roles.