The pineal gland is a dorsal median projection of the brain, laying between the telencephalic hemispheres anteriorly and the mesencephalic optic lobes posteriorly and can be associated with other outgrowths of the dorsal diencephalon (Fig. 1B’, 2F). The leatherback marine turtle Dermochelys coriacea has as adult a dorsal expansion of the endocranial cavity (* in Fig. 4C: note, the section is not complete median and the cavity is here filled with cartilage), which houses a very elongated pineal gland (Wyneken 2001: fig. 192). Paulina-Carabajal et al. (2013) interpreted this pineal cavity (partly) as the endocast rider that was first described for two extinct marine († Bothremys cooki et barberi) and one extant [Erymnochelys (“Podocnemis”) madagascariensis] pleurodire turtle (Gaffney and Zangerl 1968).

Like D. coriacea, other extant marine turtles (Chelonioidea) also have a relatively long pineal gland, which is, however, not surrounded by bone in a separate dorsal cavity (Fig. 1Q; (Wyneken 2001: fig. 193–194, 196–200; discussed by Paulina-Carabajal et al. 2017), making the pineal-gland-hypothesis very unlikely. Compared to other extant marine turtles, however, the adult leatherback has a particularly shortened and highly domed skull (Nick 1912; van Bemmelen 1896; Wegner 1959) with several related characteristics (Fig. 4C–D). The jaw musculature, for example, extends in a straight line between the skull roof and the lower jaw in adults (Schumacher 1972). Consequently, the trochlear mechanism at the otic capsule, usually present in cryptodires (Ferreira et al. 2020; Ferreira and Werneburg 2019; Werneburg 2012), was lost in this species (Burne 1905; Poglayen-Neuwall 1953; Schumacher 1972; Werneburg 2013b). The proportional skull changes in D. coriacea, in our opinion, are also mirrored in the orientation of internal skull structures, the anatomy of which can only be understood from an ontogenetic perspective and can, hence, help interpreting the origin of the pineal cavity in this species. That is, in a hatchling studied by us the pineal gland is clearly placed below the anterior tectal cartilage, and no pineal cavity is yet formed (Fig. 1K).

In the studied late term embryos, there is – with the notable exception of Chelodina with its flattened skull (Fig. 1H–5, 2I) – always a conspicuous distance between brain tissue and the surrounding skull elements, which is filled with cerebrospinal liquid. Blood vessels have enough space within this liquid (Fig. 2F) and do not leave any imprint to the internal surface of the skull in turtles that could represent a rider. In the stained CT-scans, we never identified any imprinting vasculature on the dorsal and lateral braincase wall. However, there is clear evidence that there is, in most cases, much space between brain tissue and the borders of the brain-cavity in adult turtles also (Edinger 1929; Evers et al. 2019; Ferreira et al. in press; Wyneken 2001; Zangerl 1960). Only in few specimens, such as in the hatchling of Podocnemis erythrocephala studied herein, brain tissue comes in close contact to the skull in the cerebellar region (Fig. 1B; see also Fig. 1). Podocnemis, however, shows a typically elongated (cartilaginous) rider in the endocast as is also known for specimens with a lot of space between brain tissue and the border of the brain-cavity. A hypothesis that the “rider” on the endocast surface is exclusively caused by blood vessels in all turtles must, hence, be refuted.

Although we never found any indication of the following development (maybe partly also because of limited data), one could argue that later in ontogeny the brain could – completely or partly – expand such way that it pushes the blood vessels against the braincase internally. However, the skull elements surrounding the brain, the cartilaginous chondrocranium and the dermal skull roof bones – parietal and frontal mainly –, are already well-developed around hatching (Fig. 2D–F, I) and a vascular imprint to the braincase would require enormous forces or reorganizations of the developing skull, which is not conceivable, particularly when considering the soft nature of the vessels. Also, the brain cavity apparently grows faster than the brain itself, resulting in larger endocranial spaces in older individuals (Ferreira et al. in press). Moreover, in case the soft-tissued vessels are attached to the inner wall of the skull, the propability of vessels to fossilize and then to leave an imprint on the endocast must be considered a highly exceptional event and such fossilized vessels on the skull bones were never recorded or described in any digital reconstruction in turtles.

When considering the blood-vessel hypothesis, Paulina-Carabajal et al. (2013) referred to Witmer et al. (2008), who described blood vessels on the surface of the endocast in Crocodylus johnstoni. However, the related CT-based reconstruction was done on a postmortem crocodile in which air “fortuitously had entered the encephalic venous system” (p. 72), which cannot be taken as a hard evidence to discuss the origin of the “rider”. Usually, the blood vessels collapse after death making fossilization even more unlikely. Nevertheless, the correspondence particularly of the sphenoparietal sinus of the crocodile to the v-shaped “rider” and its ventrolateral extension towards the trigeminal ganglion, as interpreted as endocast structures for † Plesiochelys etalloni (Fig. 1F–F’) by Paulina-Carabajal et al. (2013), is stunning, but appears to be a reconstruction artifact based on a scan available to us. It might be possible that a very elongated rider as seen in † Pl. etalloni might correspond to the dorsal longitudinal sinus along the skull roof (sensu Witmer et al. 2008). Nevertheless, we found no such correspondence in the sampled histological sections or µCT data, and, hence, consider this hypothesis unlikely (i.e., it rather represents the suture between the frontals). In most cases, the elongated rider might be too massive to correspond to a vessel rather than to the anteriorly extented anterior process of tectum cranii like in Chelydra serpentina (Fig. 2D–F).

Late term Ca. caretta embryos (Kuratani 1999) and hatchling Podocnemis unifilis specimens (Sheil and Zaharewicz 2014) have a very curious shape of the anterior process of the tectum synoticum with anterolateral extensions giving the process a T-shaped appearance in dorsal view (Fig. 3Ad’–B). These observations suggest that among turtles, the adult tectum cranii might develop curious shapes, including anteriorly pointing and converging branches depending on the specific skull architecture and morphofunctional requirements in the adult. It might even make connections to elements of the primary braincase wall (Fig. 2G–H), including taenia marginalis or pila antotica (Paluh and Sheil 2013), leaving traces on the lateral aspect of the rider (see black labels Fig. 1E–F as possible alternative to interpret the structures found in † Pl. etalloni).

In the late embryonic skull of turtles, the chondocranium is well differentiated and already possesses some endochondral ossifications (visible in Fig. 2C–I, 3B, N–R’). In the primordial skull, the otic capsules are dorsally connected via a tectum synoticum (Fig. 3Ad’), which dorsally borders the foramen magnum (Fig. 3C’). In its anterior part, it is dorsolaterally covered by the parietals and usually ossifies endochondrally as the supraoccipital (e.g., Fig. 3B/B’) (Sheil 2002). This cartilaginous bridge forms an anterior median process in all turtle embryos, which can be partly embedded in the ventral surface of the parietals and it projects into the brain-cavity (Fig. 2–3). The imprint of this anterior process of tectum synoticum to the skull roof bones clearly represents the anatomical correlate to the “rider” of the brain-endocast in late embryonic turtles. We never found the process to imprint the actual brain in the embryos, although it can closely align to its surface in late term specimens (Fig. 2F).

There is a striking diversity in the relative length and orientation of this late embryonic process in different species. This is, first of all, related to the developmental age of the embryos as visible in Eretmochelys imbricata (Fig. 3H vs. 3I) (Fuchs 1915; Sheil 2013) and Chelydra serpentina (Fig. 2D–F vs. 3D) (Sheil and Greenbaum 2005; see also Rieppel 1976, 1993). The older the embryo, the longer the process appears to be. Second, there are taxonomic differences among species at late embryonic stage (e.g., Fig. 3F vs. 3G). Whether, to which degree, and at which time of development the process will be replaced by endochondral ossification, and whether it will continue to grow through post-hatching development or remain as small cartilaginous process cannot be evaluated herein and is certainly different among species. It is clear, though, that at least in some taxa the most anterior part of this process remains cartilaginous after hatching, as confirmed by the µCT images of PTA-stained specimens (Fig. 1K–L, O, Q). The tectum usually ossifies from posterior to anterior (Werneburg and Yaryhin 2019) suggesting that the extension of the rider might be influenced by the completeness of ossification of the supraoccipital.

In this context, little research has been done on cartilaginous structures in post-hatching, juvenile, and adult turtle skulls. It is known, however, that at least in the marine turtles Dermochelys coriacea (Fig. 4C) and Chelonia mydas (Chelonioidea) (Fig. 4A, 3P–R’) and in the snapping turtle Chelydra serpentina (Chelydridae) (Fig. 4E) the embryonic tectum synoticum partly persists as the cartilaginous tectum cranii in the adult (Nick, 1912; Parker 1880; Wyneken, 2001: labeled as “cartilaginous part of brain case” in fig. 201–202), whereas the posterior part of the tectum is ossified as supraoccipital. This is also confirmed by our observations of the µCT images of PTA-stained specimens (Ferreira et al. in press), which clearly show that, at least in some taxa, the most anterior region of the supraoccipital is cartilaginous in juvenile and small adults (Fig. 1K–L, O, Q). The anterior process of the cranial tectum, ventrally embedded between the parietals, spatially corresponds to the rider protuberance on the endocast. In D. coriacea, Chelo. mydas, and Chely. serpentina, the proportions of the late term embryonic and the juvenile/adult cartilaginous process are – in relation to the rest of the skull – relatively similar. To which degree in the adult the cartilaginous process is calcified cannot be evaluated herein. In a juvenile specimen of Pelomedusa subrufa with a carapace length of 9.6 cm, one of the largest turtles ever treated with histological methodology (pers. comm. Wolfgang Maier), we still found a long cartilaginous process in place, whereas the rest of the skull is well ossified (Fig. 2A–B, G–H).

Considering the actual presence of a dorsally covered bony or cartilaginous anterior part of the supraoccipital, the anterior process might have some relation to the architecture of the temporal region (sensu Joyce 2007), which could be tested in future biomechanical analyses. The anterior process of the tectum synoticum is an ancestral feature also present in other reptiles (Howes and Swinnerton 1901; Werneburg and Yaryhin 2019; Yaryhin and Werneburg 2018) and serves as a general anchor (Pitirri et al. 2020) and potential force buffer for dermal roof bones.

Originally, in early Testudinata, such as † Proganochelys quenstedtii (Gaffney 1990), the temporal skull region was almost completely covered by temporal bones (Werneburg 2012; Abel and Werneburg in press). With the emergence and increase of neck retraction through turtle evolution (Werneburg et al. 2015a; Werneburg et al. 2015b), marginal reductions of the dermatocranial skull armor evolved to buffer neck muscle forces during neck retraction (Werneburg 2015).

As there is variation in rider shape among turtles, and the tectum might serve as anchor and/or buffer for biomechanical forces, some aspects of rider diversity roughly seem to be consistent with variation in skull roof emargination with very short, bulbus-like riders being associated with no or only a shallow occiput emargination and elongated riders associated with deep occiput or complete emargination. This relation is tentative and needs rigorous quantitative examination beyond the scope of this study. If rider shape and length were ultimately found to be correlated with emargination, this would reinforce current hypotheses on how important neck functionality and associated characters are for skull disparity and development in turtles.

We speculate that forces transmitted along the reduced temporal coverage during neck retraction are buffered along the postorbital area in a posterodorsal and medial direction into the anterior process of the tectum synoticum (red arrows in Fig. 4B/B’, D/D’, F/F’). A shorter process in marine turtles with shallow emargination such as Chelonia mydas may compensate forces over the skull more posteriorly (Fig. 4B, B’), whereas a longer process in taxa with deep posterior emargination such as Chelydra serpentina may compensate neck forces more anteriorly over the skull (Fig. 4 F/F’). This tentative association, however, may not be very distinct in future biomechanical analyses given that other factors certainly affect the temporal region and the skull roof, such as different skull proportions, jaw muscle action, and growth patterns.

The shape of the adult endocast rider cannot be easily associated to the length of the anterior tectal process (see Table 1; Werneburg et al. 2021), because of different modes of fossilization in extinct taxa and/or different degrees of ossification in the anterior tectal process. Nevertheless, it is noticeable that some stem turtles with their full temporal coverage, such as † Pr. quenstedtii (Lautenschlager et al. 2018) or † Meiolania platyceps (Paulina-Carabajal et al. 2017), only show a bulbus-like appearance of the rider, which might correspond to a short tectal process. Other stem turtles with deep occiput emarginations, such as the baenid paracryptodire † Eubaena cephalica (Evers and Benson 2019; Joyce et al. 2016; Rollot et al. 2018) can show elongated riders indicating to an elongated anterior tectal process.

The collected images on the late embryonic stages from the literature (Fig. 3) are not fully comparable among each other because first, they are just random snapshots on ontogenetic development and the length of the anterior process might grow quickly through ontogeny (Fig. 3A). Second, some of the images are based on reliable histology-based 3D-reconstructions (e.g., Fuchs 1915), but most of the images are based on the more critical clearing and staining methodology, which has fundamental impact on the exposure of small and thin cartilage and relative positioning of structures inside the partly enzymatically digested embryo (discussed by Yaryhin and Werneburg 2017). Nevertheless, despite those limitations, the relative lengths of the process in most cases fit relatively well to the respective emargination types (compare to Fig. 1H): A short embryonic process is generally associated to an absent or a shallow posterior emargination (1–3 in Fig. 1H) (Eretmochelys imbricata: Fig. 3H–I, Emydura subglobosa: Fig. 3F, Chelonia mydas: Fig. 3N–R’; Caretta caretta: Fig. 3Ad). A longer process is generally associated to a deep posterior emagination (4 in Fig. 1H) (Chelydra serpentina: Fig. 2D–F – note in Fig. 3D only an early stage of this species is shown, same for Macrochelys temminckii in Fig. 3J and Apalone spinifera in Fig. 3C; Podocnemis unifilis: Fig. 3B; Chrysemys picta: Fig. 3E; Emys orbicularis: Fig. 3G). Pelodiscus sinensis (Fig. 3K) and Trachemys scripta (Fig. 3M) do not fit perfectly to this categorization, which might be due to the mentioned methodological issues.

Within some extant turtle taxa, e.g. Terrapene and chelid pleurodires, a dermal coverage of the temporal region is completely lost by evolutionary expansion of the anteroventral emargination (5 in Fig. 1H) (Werneburg 2012). Compared to all other species studied herein, the histological sections of Chelodina longicollis (Fig. 2I) and the enhanced contrast stained µCT of Chelus fimbriatus (Fig. 1M) reveal an extremely broad anterior process. This suggests different functional constraints. How neck retraction forces are transmitted in species without temporal skull coverage is not understood (Werneburg 2015). Neck muscles insert to a dense temporal fascia (Werneburg 2013a), the forces within also need to be buffered. The primary function of the temporal armor was the stabilization of the mobile quadrate. Through turtle evolution, the quadrate was fixed to the braincase (Werneburg and Maier 2019) and the temporal armor was freed and could get reduced in a way to react to increased neck retraction forces but still to keep lateral bracing between quadrate and upper jaw. Without any temporal coverage, the stability of the lateral bracing was lost and the temporal fascia and the quadratojugal ligament (Jones et al. 2012; Werneburg 2013a) might not be enough to withstand forces on the skull. In this context, the very broad anterior tectal process in type 5 (5 in Fig. 1H) species (Fig. 3I) appears to stabilize the contralateral parietals to keep integrity of the skull roof (sensu Pitirri et al. 2020). We see short tectal processes and shallow emarginations in chelids and chelonoids, and long processes with deep emarginations in emydids and chelydrids. This might be a pattern explained by functional, but also phylogenetic associations. Chelodina and Chelus, with their broad process and complete (Chelodina) or peculiar (Chelus) emargination, seem to support the functional scenario, but these are only two data points. However, we present only a tentative causal mechanism for the functional association, which can be explicitly tested by biomechanical analyses in the future.