The specimen comes from the Middle Turonian of the Goulmima region, South East Morocco (Kennedy et al., 2008) and is housed at the Natural History Museum in London (NHM P.66857). This region yields abundant and exceptionally well-preserved fossils from at least two levels of concretions in a marl unit within an otherwise shallow water to peritidal carbonate succession (Cavin and Dutheil, 1999). Most of the concretions contain ammonites, occasionally teleost fish and rarely marine reptiles. Skeletal remains of batoids are present but rare (Cavin et al., 2010), and this specimen represents one of the most completely preserved chondrichthyans known from the Mesozoic.

The specimen was preserved in a hard carbonate concretion (Fig. 1) and the ventral surface of the specimen was partially exposed. Mechanical preparation using needles and pneumatic tools expose and reveal large portions of the ventral surface of the specimen. The specimen was subsequently placed in a bath of buffered 7.5% volume/volume formic acid at a pH of 3.3. The reaction was halted after only a few minutes because it proceeded more rapidly than expected when compared with similarly preserved bony fish from the same locality. This exposed a further portion of the ventral surface, including in situ teeth and dermal denticles (Fig. 2). Acid preparation also detached some prismatic cartilage tesserae from the exposed skeleton and its use was suspended in favour of more sensitive preparation techniques.

Small areas, particularly the rostrum and the olfactory capsules, were cleared of matrix using a combination of acid and mechanical preparation. Buffered formic acid was applied locally, in single drops by pipette and neutralised by 10% weight/volume sodium carbonate solution. Partially prepared areas of cartilage, thus exposed, were cleaned of matrix using a mounted needle under a binocular microscope before the cycle was repeated.

When further acid and mechanical preparation were no longer practical, computed tomography methods were used to further examine this specimen. The specimen was scanned at the High-Resolution Computed Tomography Laboratory at The University of Texas of Austin (UTCT) using their highest energy scanner. This particular scanner was used because it exceeds details attainable by conventional medical scanners. Also, it is capable of scanning large specimens of up to 30 cm in diameter. This specimen was not a candidate for neutron scans, because too much natural background radiation was detected during preliminary assessment at the Helmholtz-Zentrum Berlin, in Berlin, Germany (Hilger et al., 2006). Background radiation due to trace levels of uranium (Parker and Toots, 1970) is often present in fossilized specimens. During neutron scans, atoms become temporarily radioactive and must go through a decay period before they may be safely returned. With background levels marginally high, it was deemed unadvisable to scan.

The ultimate scan was conducted at UTCT using the P250D detector set to a small spot size, at 450 kV and 1.3 mA. There was no filter and the specimen was not wedged in any material, although care was taken during pre-scan mounting, because the specimen was partially exposed. The scan was conducted with the long axis of the body perpendicular to the scanning plane. This allowed for maximum transmission of X-rays and maximum number of cross-sectional slices generated during reconstruction.

Slice data were further analyzed using VGStudio MAX 2.0 and Adobe Photoshop in the University of Texas Digital Methods Lab. Slices were imported into VGStudio MAX 2.0 using normal ramp settings on unaltered cross-sectional slice data, using the inverse ramp setting on unaltered cross-sectional slide data, and using normal ramp settings on cross-sectional slices previously inverted in Adobe Photoshop. Once in VGStudio MAX 2.0, skeletal material was digitally prepared from the matrix using contrast thresholding, opacity optimization, and region segmentation.

The use of computed tomography was highly successful. Reconstruction resulted in 906 cross-sectional slices, each 0.25 mm thick. Surface skeletal morphology was partially sacrificed during post-scan processing in all methods in order to evaluate internal morphology. The exposed skeletal material was best examined on the actual specimen.

Initial cross-sectional images had a negative contrast of skeletal element to matrix. That is, the cartilage skeleton was less X-ray attenuating, and therefore darker than the surrounding matrix (Fig. 3A). This complicated modelling in VGStudio Max 2.0 because contrast thresholding selects for lighter voxels on greyscale images. As a result of this negative contrast, we could only render a surface model when working with unaltered cross-sectional images, which was little different than examining the specimen as it was (Fig. 3B, C).

When cross-sectional images were imported using the inverse ramp option, the matrix was darker than the skeletal cartilage, however, the entire specimen was surrounded by a mass of white space (Fig. 4A). It was possible to subtract much of the white space from the model (Fig. 4B, C) with the exception of a thin layer around the specimen. This thin layer of high contrasting voxels was likely due to beam hardening (Ketcham and Carlson, 2001). The most detail for these models was found when the background was turned black, because the model was slightly translucent (Fig. 4C).

Finally, when we inverted cross-sectional images in Adobe Photoshop and adjusted contrast levels prior to opening them in VGStudio Max 2.0, the specimen rendered relatively opaque. The Adobe Photoshop procedure left a much brighter layer of voxels that needed to be segmented away from the model (Fig. 5A). This was conducted by two methods of segmentation. The first used the region-draw tool, to select the bright layer of voxels on the dorsal surface of the specimen manually (Fig. 5B). The second used the 3D propagation tool to select the ‘blank space’ around the specimen. Once selected, the dilate tool was implemented to expand the region uniformly, to include the bright layer that surrounded the specimen (Fig. 5C).

The specimen is relatively complete in the unexposed areas of the concretion. There is a natural fracture at the approximate joint between the craniovertebral articulation. A second fracture traverses the right side of the skull. A section of the abdominal vertebral column was disarticulated from the axis at an intervertebral disc, probably post-mortem, and is shifted anteriorly, so that it runs parallel to the main vertebral axis between the pectoral girdle and pelvic girdle. The exposed skeletal elements are composed of a layer of small tesserae, prismatic cartilage, which surrounds an inner region of uncalcified cartilage. This inner region is of the same X-ray attenuation as the surrounding matrix. Without petrographic thin sections, we cannot say for certain if that is due to primary cementation or secondary replacement. Tesserae are visible around all skeletal elements, even in 3D models of elements obscured by matrix. This contributes to the hollow appearance of skeletal elements in cross-section (Fig. 3, Fig. 4 and Fig. 5) and volume rendered models (Fig. 4C). In volume models, the tesserae layer is rendered as most X-ray attenuating along the median crest of the synarcual cartilage and around the posterior portion of the chondrocranium. The areolar cartilage of the vertebral centra renders the brightest of all aspects of the skeleton and reflects its denser calcification nature. A great deal of the chondrocranium and pectoral skeleton appears weakly calcified, especially the anteriormost portion of the rostrum and the pterygial cartilages. This is consistent with the typical pattern of cartilage density observed in volume rendered models of extant elasmobranch taxa.

Three denticle morphologies, which are all similar to the denticles of the guitarfish Aptychotrema, were recognized (Fig. 2A–C). During acid preparation, an articulated section of a ventral covering of polygonal denticles was exposed and it is possible that there would have been a fairly complete covering of the ventral body with denticles. It is more common to have denticles on the dorsal surface and a naked ventral surface in extant taxa. The teeth are of underived “rhinobatid” form (Fig. 2D). The shape of the central uvula and absence of well-formed lateral uvulae are similar to those of Platyrhinoides. This is different from other modern rhinobatids and Platyrhina, but is more developed than in Zanobatus. Although the overall shape of the teeth is very different, the uvulae are not unlike those of Rhynchobatus and pristids. The largest teeth barely exceed 0.5 mm and the smaller ones, which are the most abundant, are under 0.35 mm. The root is variable with a wide groove in some, closed over in others.

Compared to other extinct and extant guitarfishes and skates, this specimen possesses a short rostral cartilage and small precerebral fontanelle, the nasal capsule is large and anteriorly directed and the mandibular arch cartilages are slender and laterally elongate. The synarcual cartilage is narrow in dorsal view, and in cross-section, the lateral stays are strongly curved anterolaterally. The median crest is wide and tapers slightly at the anterior aspect. The first vertebral centrum of the synarcual is positioned far posteriorly. The pectoral girdle is loosely connected to the vertebral column and the suprascapula is dislocated and fragmentary. The scapulocoracoid comprises a gracile, fused, coracoid bar and a shallow and rectangular scapular process. A robust propterygium is preserved proximally with a few thick proximal radials in articulation. The mesopterygia and metapterygia are difficult to discern in the fossil.

By combining traditional preparation with high-resolution X-ray computed tomography and digital preparation, we were able to produce imagery for this unique specimen that can serve as a voucher during comparative analysis. The cross-sectional images generated correspond closely to serial sections of modern taxa and when those cross-sections are compiled, they produce 3-dimensional representations that match the original specimen (Fig. 3). These data can be digitally altered by using image processing packages like VGStudio Max 2.0 or Adobe Photoshop to enhance efficiency when visualizing the skeleton (Fig. 4 and Fig. 5). By performing digital preparation, we could maintain the integrity of the fossil and get an accurate representation of the relative position of skeletal elements. In our opinion, the results of using the invert ramp option in VGStudio Max 2.0 produced the best results to see the complete skeleton of this specimen (Fig. 4C).

Three-dimensional preservation of chondrichthyans is rare, but under the right circumstances, concretions may preserve a skeleton close to its original shape. This specimen shows exceptional detail for a fossil elasmobranch and is potentially only the second 3-dimensionally preserved fossil batoid taxon discovered. The first 3D Cretaceous batoid was †Iansan, Brito and Seret, 1993. Our results indicate that this specimen differs substantially from †Iansan, as well as most other Mesozoic batoids and modern taxa such as Rhinobatos. It is most like members of the Platyrhinidae, and may prove instrumental in understanding the radiation of that lineage.