Description. BMNH R2095 (Text-figs 3–4) is a partial dorsal vertebra from the middle or posterior portion of the dorsal column. Most of the centrum and neural arch are preserved, but the condyle is broken, and the neural spine and dorsal part of the neural arch are missing, as are the pre- and postzygapophyses and diapophyses. However, sufficient laminae remain to allow the positions of the processes to be inferred with some certainty (Text-fig. 5). Measurements are summarized in Table 1.
The most striking features of this specimen are the extreme height, anteroposterior length and anterodorsal inclination of the neural arch. These are clearly genuine osteological features and not the result of post-mortem distortion. Although the dorsalmost preserved part of the neural arch is ventral to the diapophyses, the height even of the remaining portion (160 mm above the anterodorsal margin of the centrum, measured perpendicular to the anteroposterior axis of the centrum) is equal to that of the cotyle. The centrum is 190 mm long measured along its dorsal margin; its anteroventral portion is missing but a maximum length of 200 mm is indicated, assuming that the curvature of the condyle is approximately equal to that of the cotyle. The base of the neural arch is 170 mm in anteroposterior length, 85 per cent of the estimated total length of the centrum, and its posterior margin is continuous with that of the cotyle, forming a single smooth curve when viewed laterally. The angle of the neural arch's inclination relative to the vertical cannot be precisely ascertained due to the absence of the condyle, but was approximately 35 degrees and cannot have varied from this by more than 5 degrees or so unless the condyle was shaped very differently from that of other sauropods.
A clean break of the condyle exposes within the centrum the dorsal part of a median septum and a pair of ventromedially directed lateral septa, indicative of an extensively pneumatized centrum with camerate, rather than camellate to somphospondylous, internal structure. The ventral portion of the broken condyle cannot be described as it is obscured by a catalogue note. The cotyle is slightly concave, its central portion indented 10–15 mm relative to its margin. It is 160 mm tall and 170 mm wide. A very subtle keel is present on the ventral surface of the centrum, and the ventral border of the centrum is gently arched in lateral view.
On the better preserved left side of the vertebra, a shallow lateral fossa is positioned dorsally on the centrum, and about midway between the anterior and posterior margins of the neural arch, onto which it intrudes. It is very roughly triangular in shape, taller anteriorly than posteriorly, with a maximum height of 80 mm and a total length of 95 mm. Set within this is a deeper lateral foramen, oval, anteroposteriorly elongate and measuring 80 by 40 mm. The fossa and foramen share their ventral borders. On the right side, the lateral fossa is situated even more dorsally, but is taller posteriorly than anteriorly, with a maximum height of 55 mm and a total length of 90 mm. The lateral foramen is much smaller on this side, measuring only 20 by 15 mm, and is anteroventrally placed within the fossa.
On the left side, the dorsal border of the lateral fossa is formed by a prominent sharp-lipped lateral ridge, which extends anterodorsally for 90 mm; this is absent on the right side, apparently due not to damage but to intravertebral variation. Instead, an irregularly shaped and sharp-lipped border separates the fossa from a more dorsally placed subcircular ‘accessory fossa’ 30 mm in diameter. On this side, an accessory lamina connects the anterior part of the border between the main and accessory fossae to a prominent boss positioned on the anterior margin of the neural arch, 50 mm above the anterodorsal margin of the centrum. This is not a parapophysis or a diapophysis but seems to be an aberrant feature of this individual. Neither the accessory fossa nor the anterior boss has been reported in any other sauropod vertebra; however, these features are not considered taxonomically significant as their occurrence on only one side of the vertebra suggests that they are either pathological or a developmental aberration. Pneumatic features vary wildly and may be opportunistic, if Witmer (1997, p. 64) is correct that ‘Pneumatic diverticula are … opportunistic pneumatizing machines, resorbing as much bone as possible within the constraints imposed by local biomechanical loading regimes.’
The remaining features are described from the left side of the vertebra. The right side is consistent with this morphology, although not all features are preserved.
From a point anterior to the anterodorsal margin of the lateral fossa, a vertically orientated ACPL extends dorsally 70 mm to a cross-shaped junction of laminae near the anterior margin of the arch, and may also have extended a similar distance ventrally although damage makes it impossible to establish this. The cross-shaped junction is interpreted as the location of the parapophysis. In sauropods, the position of the parapophysis migrates dorsally in successive dorsal vertebrae, being located ventrally on the centrum of anterior dorsals, dorsally on the centrum in mid to anterior dorsals, and on the neural arch of mid to posterior dorsals, level with the prezygapophyses in the most posterior dorsals: see, for example, Hatcher (1901, pl. 7). The high position of the parapophysis on the neural arch of R2095 indicates a mid to posterior placement of the vertebra within the dorsal column, but, because the prezygapophyses must have been dorsal to it, it was probably not among the most posterior vertebrae in the sequence.
In addition to the ACPL, three further laminae radiate from the parapophysis: part of an anteriorly directed PRPL, the ventral portion of a dorsally directed lamina, which is interpreted as a PPDL, and a posteroventrally directed accessory lamina supporting the parapophysis. This is presumably homologous with a PCPL, but cannot be so named as it does not approach the centrum, and indeed extends only 30 mm. Where the latter lamina merges with the neural arch, another accessory lamina arises. Directed posterodorsally, it presumably extended to the postzygapophysis and is here regarded as an accessory postzygapophyseal lamina similar to that found in posterior dorsal vertebrae of Diplodocus carnegiiHatcher, 1901 (Hatcher 1901, pl. 7). The PPDL, accessory infraparapophyseal and accessory postzygapophyseal lamina form three sides of a quadrilateral fossa; the fourth side, presumably formed by a PODL, is not preserved, although a very low and unobtrusive accessory lamina does join the dorsalmost preserved part of the PPDL to the accessory postzygapophyseal lamina. The near-vertical orientation of the PPDL indicates that the diapophysis was located some distance directly dorsal to the parapophysis, further extending the inferred height of the neural arch and ruling out an interpretation of the accessory postzygapophyseal lamina as the ACDL or as the ‘accessory PCDL’ of Salgado et al. (2005). Finally, a broken ridge of bone extends up the posterior margin of the lateral face of the neural arch. Its identity is problematic: it cannot be a PCDL owing to the anterior position inferred for the diapophysis.
Between the ACPL and the posterior lamina, above the dorsal margin of the lateral fossa and below the accessory laminae described above, the lateral face of the neural arch is a flat featureless area measuring 90 mm anteroposteriorly and 50 mm dorsoventrally. This feature is not observed in any other sauropod vertebra.
In posterior view, the pedicels of the neural arch are robust pillars, leaning somewhat medially, measuring 30 mm in width, extending at least 130 mm dorsally, and merging into the CPOLs before damage obscures their further extent. They enclose a neural canal that is almost exactly circular, 35 mm in diameter. There is no trace of the postzygapophyses or hyposphene, and no indication that these structures were attached to the preserved portion of the arch. It must be assumed, then, that these features were located on the lost, more dorsal, part of the neural arch. The hyposphene, if present, was located at least 90 mm dorsal to the centrum (measured from the floor of the neural canal), and the postzygapophyses at least 140 mm dorsal to the centrum.
In anterior view, too, the pedicels are robust, being 25 mm in width. They merge gradually into the CPRLs and extend dorsally for at least 80 mm, dorsal to which they are broken. In this aspect, however, the neural canal has no roof, instead forming a large teardrop-shaped vacuity 120 mm tall and 55 mm wide. The dorsal portion of this vacuity is bounded by a pair of gently curved, dorsomedially directed laminae unknown in other sauropods, which meet at a 55 degree angle to form an arch dorsal to the neural canal. The vacuity is filled with matrix, so the extent of its penetration posteriorly into the neural arch cannot be assessed. The prezygapophyses are absent; their articular surfaces were probably about 140 mm above the floor of the neural canal, judging by the trajectory of the PRPL.
The most anterodorsal preserved portion of the vertebra is obscured by a flat, anterodorsally directed ‘apron’ of matrix, 15 mm thick and 120 mm wide, which hampers interpretation of the prezygapophyseal area.