Thin-sections were prepared from the caudal ends of five parietals representing various growth stages of Triceratops, which includes the adult form originally regarded as Torosaurus (see Horner and Goodwin, 2006 and Scannella and Horner, 2010). In addition, two isolated epiparietal bones were sectioned (see Table 1). Specimens were acquired from the palaeontology collections of the Museum of the Rockies, Bozeman, Montana, USA (MOR) and the University of California, Berkeley, California, USA (UCMP). The histological slides analyzed in this study are listed in Table 1. Sections were taken from approximately the same location on each parietal and epiparietal (see Fig. 1A), and each was cut in three orientations: sagittal, transverse and coronal (see Supplemental Information for specific descriptions of these orientations).

Many of the fossil specimens were highly porous with unfilled vascular spaces, and some fossils were very soft and fibrous; readily absorbing water even following glue-consolidation and resin-embedding. Due to the atypical composition of this cranial material, many adaptations to standard fossil thin sectioning (Lamm, 2007) were employed (see Supplemental Information).

The smallest parietal (Parietal 1) (MOR 1199, Fig. 1B–F) is composed exclusively of a cancellous, primary osseous tissue giving the bone a very spongy appearance (Fig. 1B). Although some of the vascular spaces are relatively large, there is no evidence that osteoclasts were involved in expanding the channels, as the borders of these channels are not scalloped. In sagittal section there is an overall radial pattern to the vascular spaces (Fig. 1C), but in coronal view this radial pattern is less distinct (Fig. 1D). In transverse section (Fig. 1E, F), flattened osteocyte lacunae encircling some of the vascular canals suggest the presence of an endosteal centripetal deposit, although upon close inspection (Fig. 1F), there is no evidence of a cementing line, and little evidence of lamellar tissues. Most of the bone has a simple woven texture. The encircling tissues are apparently in the initial-most stages of forming primary osteons. There is no evidence of any resorption or secondary reconstruction.

The smallest epiparietal (Epiparietal 1) (UCMP 159233, Fig. 1G–I), measuring two centimeters in height, is very porous, and composed of what appears to be a primary tissue (Fig. 1H), but lacks structures identifiable as osteocyte lacunae. In transverse section (Fig. 1I) each of the vascular channels is simple with no evidence of lamellar tissues. An overall globular appearance of the matrix gives the impression that the section has transverse bundles of fibers, but fibers are not evident in the longitudinal (sagittal) section. Dark areas in the slide apparently represent regions of microbial invasion or possibly a mineral precipitate.

The next larger size parietal (Parietal 2) (MOR 2969, Fig. 2A–F) is nearly twice the overall dimensions of the first, and more than twice its dorso-ventral thickness (Fig. 2A). Even though the bone is much larger, the section looks very similar to the smaller individual. The tissue has a very spongy appearance with many radial canals. In coronal section, (Fig. 2B, C) the vascular spaces are large (Fig. 2B), osteocyte lacunae are abundant, and all the tissue is primary with a much woven texture (Fig. 2C). In transverse section, (Fig. 2D–F) one of the large openings is obviously expanded by osteoclasts as evidenced by the crosscutting of primary osteons (Fig. 2D, arrows), but these osteoclastic enlarged spaces are not common. Primary osteons are evident with flattened lacunae encircling the vascular canals (Fig. 2E). Upon magnification, the matrix surrounding many of these vascular spaces is highly fibrous (Fig. 2F).

An isolated epiparietal (Epiparietal 2) (MOR 3060, Fig. 3A–D) is the size and shape (see Horner and Goodwin, 2006 and Horner and Goodwin, 2008) to adorn a parietal frill the size of MOR 2969 (Fig. 2) and has a spongy central region surrounded by a denser, less vascularized exterior (Fig. 3A). In sagittal view the vascularity has a rostral-caudal orientation, and is composed of secondarily reconstructed tissues, as seen in polarized light (Fig. 3B). Near the exterior surface (Fig. 3C) the tissue is very fibrous and contains no evidence of osteocyte lacunae. In transverse section (Fig. 3D), the magnified matrix appears globular where the slide transects bundles of fibers.

The third parietal (Parietal 3) (MOR 3027, Fig. 4A–G) in this study reveals some very interesting differences from Parietals 1 and 2. Although Parietal 3 is fifteen centimeters longer than Parietal 2, it is actually dorso-ventrally thinner (Fig. 4A). The majority of the section reveals primary tissues, particularly at the caudal end (Fig. 4B, C), but there is an ontogenetic progression of tissues between the caudal and rostral ends of the sample (contrast the tissues in Fig. 4C [caudal] with those in Fig. 4F). For this reason we have identified the location of the sections that reveal the exterior surface in the image (see letters around Fig. 4A). The section in Fig. 4G was taken from the center of the sample, between locations D and E.

Tissues located at the caudal end (locations B and location C of Fig. 4A) are primarily woven tissues with simple vascular canals. There is fountain-like pattern to the vascular canals as they initially radiate caudally (centrally), but then fan out toward the dorsal and ventral surfaces (Fig. 4B). Hook-like projections point ventrally along the rounded dorso-caudal and ventro-caudal edges (Fig. 4B, C). Resorption (arrows) on both the dorsal and ventral surfaces is evidenced by deep gouges and eroded primary osteons in the mid-section of the sample (Fig. 4D, E, transverse sections). Erosion of secondary osteons is evidenced at the rostral end of the sample (Fig. 4F). Near the center of the sample, between locations D and E (see Fig. 4A), the majority of the tissue is primary (Fig. 4G, transverse section) but protruding through this primary tissue are secondary osteons oriented perpendicular (forming the yellow “Z”) to the rostral-caudal length.

Parietal 4 (MOR 2923, Fig. 5A–E), which includes a fused epiparietal (Fig. 6A–F) is derived from a two-meter long skull with a parietal about ninety centimeters in length. Although the epiparietal is fused to the parietal at this stage of ontogeny (Fig. 5A) certain characteristics allow us to determine the delineation of the two elements (Fig. 5B). Primarily we know that the large cavern on the ventral side (left side) represents a channel for a large blood vessel that enters each epiparietal and episquamosal from the ventral side of the parietal. The bases of these vessel channels are indented into the surface of the parietal and squamosals, and form caverns in the epiossifications (Scannella and Horner, 2010).

At this stage the histology of the parietal has changed dramatically in that the tissues are either highly fibrous (Fig. 5C, D) or secondarily remodeled (Fig. 5E). Fiber orientation is not obvious as there are juxtaposing orientations in all regions of the section. In some areas fibers seem to be truncated by vascular openings (center of Fig. 5C), or in some cases the vascular spaces appear to be lined with fiber bundles (central spaces in Fig. 5D). Regardless of whether it is capped by tissues of the epiparietal, the fibrous regions are all located near the external surface of the parietal, which is being resorbed, as evidenced by truncated fibers. This unorganized, highly fibrous tissue is identical to the tissues described as having been mineralized from connective tissues through the process of metaplasia (Main et al., 2005, Ricqlès et al., 2001 and Vickaryous and Hall, 2008).

The fused epiparietal is primarily composed of secondarily remodeled tissues (Fig. 6A), but a closer inspection, particularly on the dorsal and ventral sides, reveals a zone of highly fibrous tissue with non-oriented fibers (Fig. 6B–F). At the apex of the epiparietal secondary osteons are being resorbed (Fig. 6A). Deeper into the fibrous areas some of the vascular spaces seem to truncate fibers (Fig. 6C, D), and in some cases have scalloped walls (arrow in Fig. 6C). The birefringent pattern in Fig. 6D reflects the various directions of the fiber bundles, and is very similar to the pattern seen in ankylosaur osteoderms (see Main et al., 2005, Fig. 7). The base (rostral-most end on the ventral side) of the epiparietal is highly fibrous with large openings (Fig. 6E), none of which appear to be excavated by osteoclasts. Instead this appears to be a “growing” zone where mineralized tissues are apparently accreting. The borders of these openings are lined with rounded bundles of fibers. As has been described previously (Horner and Goodwin, 2008), the epiparietals and episquamosals change shape considerably during ontogeny. These elements start out as tiny triangular bones that inflate into large triangular structures, and then “flatten” down against the caudal edges of the parietal. As they flatten rostrally a portion of the bone seems to be displaced down the ventral edge of the parietal, producing the sagittal shape seen in Fig. 5B (highlighted by the black line). A highly magnified view of these fibers shows no apparent evidence of osteocyte lacunae (Fig. 6F), although it is possible that the lacunae are flattened and hidden within the fibers.

Parietal 5 (MOR 981, Figs. 7A–F, 8A–F) thin-sections were taken at a location between epiparietals, and therefore lack epiparietal tissues. The caudal tip of the parietal is highly fibrous (Fig. 7B, arrow), and although there are black structures that have the general appearance of osteocyte lacunae, they lack canaliculli, and are actually spaces between fiber bundles. On the ventral side of the section near the caudal end (Fig. 7B) the fiber bundles are arranged in a pattern that resembles herringbone (Fig. 7C). At other locations rostrally, and on both the dorsal and ventral sides, dense Haversian tissues are being resorbed at the surfaces (Fig. 7D, E, arrows). In the central area of the parietal, the tissues are all secondarily reworked, but in some intervening areas between the exterior surfaces and the central core (Fig. 7F), there are regions in which secondary osteons (red), identified by encircling cement lines, reside in a primary matrix that is highly fibrous (green). The fibers are generally oriented in a rostral-caudal direction (the caudal direction is toward the upper left corner of the image).

In transverse section (Fig. 8A–D) these secondary osteons (Fig. 7F) show up as elongated birefringent structures (Fig. 8B, green). The long axes of these Haversian systems are therefore oriented laterally, parallel to the caudal edge of the parietal. The transverse section through the fibrous matrix (the green background of Fig. 7F) shows what at first glance looks like fibrolamellar tissue with embedded osteocyte lacunae and dark parallel zones (Fig. 8C), but the black banding is actually composed of layered, vertically oriented fiber bundles (white arrows). At the base of the dark layered fiber bundles are thin light colored lines that contain very little fiber (yellow arrows). A closer examination of the matrix between the dark fiber bands reveals a very globular, fibrous texture (Fig. 8D), with lacunae (black structures) in between (Fig. 8E, F). This population of lacunae vary greatly in shape and size, and the identification of their progenitor cells (osteoblasts or fibroblasts) is therefore questionable. The vascular canals are all simple.

Parietal 1 and Parietal 2 are composed of a cancellous, primary osseous tissue that contains an abundance of radially oriented vascular canals. Tissues with these characteristics are described as being the product of hyperostosis (see Francillon-Vieillot et al., 1990, page 495), and known to exist in the skulls of pachycephalosaurid dinosaurs during initial stages of growth (Francillon-Vieillot et al., 1990, Goodwin and Horner, 2004 and Reid, 1996). Quantitative studies of the relationship between growth rates and tissue types of extant birds have established that radial fibrolamellar bone tissues are deposited very rapidly (de Margerie et al., 2002; de Margerie et al., 2004). The frill not only expanded rapidly in length and width, but inflated to a dorso-ventral thickness of more than four centimeters, greater than twice the thickness of the mature adult frill. As seen in the parietals of the subadults (Parietal 3 and Parietal 4) and adult (Parietal 5), following initial inflation of the parietal, external resorption and internal compaction occurred concurrent with continued growth in length and width. The secondary osteons, oriented perpendicular to the rostral-caudal length of the parietal (Fig. 4F, G), show that secondary reconstruction (osteoclastic resorption forming internal erosion rooms) followed the widening of the parietal rather than its lengthening. We see that this process continued into adulthood (see Fig. 8B). In the final stages of frill development, when the epiparietals began to fuse to the parietal, the caudal exterior surfaces of the parietal appear to have extended in length through the addition of mineralized connective tissues, possibly formed by metaplasia (see Haines and Mohuiddin, 1968 and Vickaryous and Hall, 2008). An abundance of juxtaposing, dense fiber bundles and lack of osteocyte lacunae in the smallest epiparietals (Fig. 1G–I), suggest that these structures may have originated through metaplasia rather than having been formed by osteoblasts.

The light-colored parallel lines (yellow arrows in Fig. 8C) seen in the image showing layers of dense fibers (white arrows in Fig. 8C) near the exterior surface apparently represent “modulations” (see Main et al., 2005, p. 303) of connective tissue metaplasia as described in ankylosaurian osteoderms. Apparently, connective tissues mineralized at the surface, and as successive fiber bands mineralized, the deeper bands were reconstructed. Osteoclasts invaded forming elongated erosion rooms oriented along the rostral-caudal length of the parietal. Haversian reconstruction filled the erosion rooms.

The ontogeny of the Triceratops parietal, as interpreted from sections derived from the caudal border, suggests that the parietal frill of juveniles grew rapidly (based on the abundance of radial canals) through a process of non-pathologic hyperostosis, which created a thickened, cancellous, spongy structure. The over-thickening of the structure was probably the result of selection to produce a large, relatively robust structure in a very short period of time, possibly for status recognition (Padian and Horner, 2011). When the parietal reached its maximum thickness, bone apposition became restricted to the growing borders while exterior resorption and internal compaction dorso-ventrally thinned the element. Internal compaction and reconstruction strengthened the structure as it became thinner. External resorption apparently progressed at different rates in different regions of the parietal as evidenced by the eventual dual fenestration of the adult parietal (Scannella and Horner, 2010). As the parietal continued to extend in length and width and to thin dorso-ventrally, the structure was eventually constructed of dense Haversian bone in all regions except at the expanding borders. Final additions to the length of the parietal were made possible by the mineralization of connective tissues (apparently through a process of metaplasia), and the fusion of the epiparietals. Here we hypothesize that the final stage of parietal development was the reshaping of the fused epiparietals to form an apparent metaplastic border to the frill. This hypothesized metaplasia of connective tissues occurred in waves or “modulations” (see Main et al., 2005) evidenced by the bands of vertical fiber bundles paralleling the parietal border.

It is likely that the parietal frill of a mature adult weighed about the same as the frill of a large subadult (adult Triceratops previous to the synonomy of Torosaurus with Triceratops by Scannella and Horner, 2010), meaning that although the parietal frill expanded in length and width, its concurrent thinning and fenestration probably kept the weight stable. The highly fibrous, mineralized (metaplastic) border suggests that the outer edge of the parietal was formed of a dense connective tissue. Most of the frill, however, consisting of a thin, fenestrated, dense Haversian bone tissue would likely have been ill-suited for protection from physical combat (see Farke et al., 2009).

Interestingly, the results of this study revealed a great deal of similarity with the results of Tumarkin-Deratzian (2010), in that the ontogenetic growth of the Triceratops parietal is similar to the growth of the Centrosaurus parietal. In our study, however, we found that the third phase of growth involved mineralization of a dense, fibrous tissue, hypothesized to have formed through metaplasia, whereas Tumarkin-Deratzian suggested the third phase to involve normal lamellar tissues. The accretion illustrated by Tumarkin-Deratzian (2010: Figure 17, 4D, page 257) shows what certainly looks like lamellar bone tissue, but what in Triceratops (upon close examination) is actually dense bands of fibers, hypothesized not to have been deposited by osteoblasts, but instead, mineralized from connective tissues. Further comparative osteohistologic studies will be required to resolve whether or not metaplastic processes are involved with the final formation of the centrasaurine frill, or whether its ontogenetic histology is possibly dissimilar to that of Triceratops.

This study was preliminary in that it involved the analysis of a few ontogenetic samples from the caudal ends of Triceratops parietals. A thorough study is underway in which thin-sections are being taken from various locations on all ornamented bones at different ontogenetic stages of Triceratops. Preliminary observations of these sections suggest that growth processes are similar throughout the skull.