A specimen-level phylogenetic analysis and taxonomic revision of Diplodocidae (Dinosauria, Sauropoda)

Positional terms for vertebrae

Serial variation within the vertebral column is highly developed in sauropods and is of taxonomic importance (Wilson, 2002; Wilson, 2012). The high level of observed variability requires detailed character descriptions restricted not only to cervical, dorsal or caudal vertebrae, but even to areas within these respective portions of the column. It is thus common for phylogenetic analyses of sauropod dinosaurs to include characters that are restricted to anterior cervical vertebrae, or mid- and posterior caudal vertebrae, for example (e.g., Wilson, 2002; Upchurch, Barrett & Dodson, 2004; Upchurch, Tomida & Barrett, 2004; Whitlock, 2011a; Mannion et al., 2012; Tschopp & Mateus, 2013b). However, few papers include definitions of these subdivisions. The definitions used in the present analysis mostly follow the ones proposed by Mannion et al. (2013), and are summarized in Table 3.

Amphicoelias altus, AMNH 5764* and AMNH 5764 ext*

The holotype of Amphicoelias altus originally included a tooth, two dorsal vertebrae, a pubis, and a femur (Cope, 1877a). A scapula, coracoid, and an ulna were later provisionally referred to the specimen (Osborn & Mook, 1921). However, the strongly expanded distal end of the scapula, and the relatively deep notch anterior to the glenoid on the coracoid actually resemble more Camarasaurus than any diplodocid (McIntosh, 1990b; E Tschopp, pers. obs., 2011). The same accounts for the single tooth stored at AMNH (Osborn & Mook, 1921). The tooth has already been excluded from scores of A. altus in recent phylogenetic analyses (Whitlock, 2011a; Mannion et al., 2012), which is followed here. Mannion et al. (2012) furthermore excluded the referred forelimb elements. Given that personal observations confirmed the rather camarasaurid than diplodocid morphology of the scapula and coracoid, but not particularly the ulna, two different preliminary phylogenetic analyses were performed with a reduced (excluding the tooth, the scapula and the coracoid, but including the ulna) and the extended holotype Amphicoelias altus OTU (including all referred elements other than the tooth). Because both analyses yielded the same position for the specimens, the reduced holotype was preferred in the final analysis. The risk of adding dubious information from potentially wrongly referred material was thus circumvented. More detailed analysis is needed in order to refine these assignments.

“Amphicoelias” latus, AMNH 5765*

This is a fragmentary specimen comprising four caudal vertebrae and a right femur from the same site as the holotypes of Camarasaurus supremus and Amphicoelias altus (Cope, 1877a; Osborn & Mook, 1921; Carpenter, 2006). Both the vertebrae and the femur show greater resemblance with Camarasaurus than to Amphicoelias, which led Osborn & Mook (1921) to synonymize A. latus with C. supremus.

Apatosaurus ajax, YPM 1860*

The holotype of Apatosaurus ajax also constitutes the genoholotype of Apatosaurus (i.e., A. ajax is the type species of Apatosaurus). During collection and shipping it became intermingled with YPM 1840, the holotype of Atlantosaurus immanis (McIntosh, 1995). As a result, it is currently difficult to distinguish the two individuals, even though they come from different quarries. We follow the suggestions of Berman & McIntosh (1978) and McIntosh (1995) in deciding which elements of the mingled taxa comprise the holotype individual of Apatosaurus ajax. The only material not confidently referable to either specimen is a braincase currently labeled ‘YPM 1860.’ In order to investigate the taxonomic implications of the attribution of this braincase to the types of Apatosaurus ajax or Atlantosaurus immanis, two supplementary analyses were performed with scores of the braincase added to YPM 1840 and 1860, respectively. Adding the information from the braincase to YPM 1840, tree length increases but positions of the two specimens remain the same. An assignment of the braincase to the holotype of Apatosaurus ajax appears thus more parsimonious, supporting the possibility that it was labeled correctly.

Apatosaurus ajax, AMNH 460*

This specimen was recovered as Apatosaurus ajax in the specimen-based phylogenetic analysis of Upchurch, Tomida & Barrett (2004). AMNH 460 is currently mounted with reconstructed portions based on other specimens. Therefore, caution was used, to avoid scoring characters based on material belonging to other individuals (for a list of bones belonging to AMNH 460, see Table S1).

Apatosaurus ajax, NSMT-PV 20375

Described by Upchurch, Tomida & Barrett (2004), this specimen is the only fully described skeleton previously referred to A. ajax. It is relatively complete, although abnormal length ratios of the humerus, radius and metacarpal III suggest that NSMT-PV 20375 might be composed of more than one individual, possibly including bones of the Camarasaurus specimens found intermingled in the quarry (Upchurch, Tomida & Barrett, 2004). These forelimb elements were thus excluded from scores of the OTU in the present analysis.

Apatosaurus laticollis, YPM 1861*

Apatosaurus laticollis is based on a single, fragmentary cervical vertebra (Marsh, 1879). Subsequent studies proposed that this vertebra actually belongs to the same individual as the holotype material of Atlantosaurus immanis (YPM 1840), which were both found in the Lakes Quarry 1 (McIntosh, 1995). Here, the specimens were kept apart in order to evaluate this hypothesis.

Apatosaurus louisae, CM 3018* (holotype) and CM 11162*

The most complete specimen of Apatosaurus is CM 3018, a postcranial skeleton that was preliminarily described as a new species by Holland (1915a) and reassessed in a detailed monograph by Gilmore (1936). An obvious diplodocid skull (CM 11162) was found near it, but the referral of this skull remained confused for a long time (Holland, 1915b; Holland, 1924; Berman & McIntosh, 1978). Because Apatosaurus was thought to have a short, Camarasaurus-like skull at the time, Holland’s proposal that CM 11162 was the actual skull of CM 3018 (Holland, 1915b; Holland, 1924) was generally rejected (e.g., Gilmore, 1936). Only with the detailed description and study of the specimen by Berman & McIntosh (1978) was CM 11162 recognized as the now widely accepted long skull-form of Apatosaurus. Given the small distance between skull and postcrania in the quarry, as well as the perfectly fitting size of the cranial occipital condyle and postcranial atlas, the probability that the two belong to the same individual is very high (Holland, 1915b; Berman & McIntosh, 1978). Accordingly, the OTU representing the holotype of Apatosaurus louisae in the present analysis comprises scoring from both CM 3018 and 11162.

Apatosaurus louisae, CM 3378*

This specimen was identified as Apatosaurus louisae in the analysis of Upchurch, Tomida & Barrett (2004). Although it has never been described in detail, CM 3378 yields important information on the number of vertebrae in Apatosaurus, as this specimen is the only one known with an articulated, uninterrupted vertebral column from the mid-cervical region to the last caudal vertebra (Holland, 1915b; McIntosh, 1981). CM 3378 was found at the Dinosaur National Monument, associated with a diplodocid skull (CM 11161; interpreted as Diplodocus), as well as appendicular elements. However, according to McIntosh (1981), these materials cannot be attributed to the same individual as CM 3378 with certainty, and no scores from them were thus included in this OTU.

Apatosaurus louisae, LACM 52844*

As with other specimens previously identified as A. louisae, LACM 52844 also comes from the Dinosaur National Monument quarry. It was found nearly complete and mostly articulated, just below the holotype CM 3018 and skull CM 11162 (McIntosh & Berman, 1975; Berman & McIntosh, 1978). Originally, LACM 52844 was housed at CM and bore the accession number CM 11990 (McIntosh, 1981). Although it was reported to be nearly complete (McIntosh, 1981), only a limited number of bones were located and scored at LACM during our study (Table S1; E Tschopp, pers. obs., 2013).

“Apatosaurus” minimus, AMNH 675*

Initially described as new species of Apatosaurus (Mook, 1917), AMNH 675 is now generally considered an indeterminate sauropod, with affinities to Macronaria, based on pelvic girdle morphology (McIntosh, 1990a; Upchurch, Barrett & Dodson, 2004; Mannion et al., 2013). In order to test this, Isisaurus colberti was added to the analysis. Isisaurus has the typical titanosaurian sacrum with six vertebrae and the preacetabular lobe oriented perpendicular to the vertebral axis (Jain & Bandyopadhyay, 1997), as is the case in AMNH 675. A diplodocid chevron is also accessioned under AMNH 675. However, AMNH records indicate it was ‘found loose with other Bone Cabin Quarry material.’ We therefore excluded it from the A. minimus OTU.

Apatosaurus parvus, UW 15556

This specimen was found by the Carnegie Museum, intermingled with the holotype specimen of Elosaurus parvus, CM 566 (Hatcher, 1902; Peterson & Gilmore, 1902). It was initially accessioned as CM 563, but was later transferred to the University of Wyoming (McIntosh, 1981). Usually identified as A. excelsus (Gilmore, 1936), a specimen-based phylogenetic analysis supported the retention of the species A. parvus for CM 566 and UW 15556 (Upchurch, Tomida & Barrett, 2004).

Apatosaurus sp., BYU 1252-18531*

Only one mention of this specimen exists, discussing sacral rib anatomy (D’Emic & Wilson, 2011). It was found in Utah, and is nearly complete and largely articulated (E Tschopp, pers. obs., 2013). The specimen is partly on display at BYU, where it is labeled as A. excelsus. No more detailed information can be given because the specimen is currently under study.

Apatosaurus sp., FMNH P25112

Riggs (1903) described this specimen (formerly FMNH 7163) as A. excelsus, which led him to two important conclusions: (1) Brontosaurus is a junior synonym of Apatosaurus, and (2) during ontogeny, additional vertebrae are added from the dorsal and caudal series to the sacrum. Later, the specimen-based phylogenetic analysis of Upchurch, Tomida & Barrett (2004) recovered it on a disparate branch within Apatosaurus, suggesting that FMNH P25112 represents a novel species. The specimen is mounted at FMNH together with the neck and forelimbs of FMNH P27021 (W Simpson, pers. comm., 2013).

Apatosaurus sp., ML 418*

This specimen is very badly preserved. It was identified as a possible Dinheirosaurus, Apatosaurus, or a yet unknown, indeterminate diplodocid (Antunes & Mateus, 2003; Mateus, 2005; Mannion et al., 2012). One dorsal vertebra has been prepared and additional unprepared material includes dorsal rib fragments, and a partial tibia. A mid- or posterior cervical vertebra of the same individual was lost due to the friable preservation, and scores concerning the cervical vertebrae are therefore based on photographs taken prior to their loss.

“Atlantosaurus” immanis, YPM 1840*

This is possibly the same individual as YPM 1861 (Apatosaurus laticollis), and it was mingled with YPM 1860 (Apatosaurus ajax) during shipping (see above). McIntosh (1995) tried to separate them based on their color, and on sparse field notes. In the YPM collections, the specimens are still labeled as they were before McIntosh’s study, therefore it is difficult to reproduce his results. Scores for an ischium of YPM 1840 are based on personal observation, whereas cervical and dorsal vertebral characters are derived from the literature (Marsh, 1896; Ostrom & McIntosh, 1966; Upchurch, Tomida & Barrett, 2004).

Australodocus bohetii, holotype* and paratype*

The holotype and paratype of Australodocus bohetii are two successive mid-cervical vertebrae from the same individual (Remes, 2007). A. bohetii was initially described as a diplodocine (Remes, 2007), but Whitlock (2011a) and Whitlock (2011c) suggested titanosauriform affinities for the species. Subsequently, Mannion et al. (2013) suggested Australodocus to be a non-lithostrotian titanosaur. Accordingly, Ligabuesaurus leanzai was added to the taxon list in order to include a possible closely related derived titanosauriform that has anatomical overlap with A. bohetii.

Barosaurus affinis, YPM 419*

The holotype of B. affinis consists only of pedal material, and has no overlap with the holotype of B. lentus (Marsh, 1890; Marsh, 1899). Because they come from the same quarry, the two species were usually regarded as synonyms (Lull, 1919; McIntosh, 2005). McIntosh (2005) identified the elements as mt I and partial mt II, but the latter is herein interpreted to represent the proximal portion of mt V instead. The bone is widely expanded, and has the typical ‘paddle’-shape of the metatarsal V in sauropods (E Tschopp, pers. obs., 2011).

Barosaurus lentus, YPM 429*

Although this specimen is the genoholotype of Barosaurus (Marsh, 1890; Lull, 1919; i.e., B. lentus is the type species of Barosaurus), most characterization of Barosaurus is based on another, more complete, and articulated specimen (AMNH 6341, see below). YPM 429 as presently available has a high degree of reconstruction, especially in some cervical vertebrae.

Barosaurus sp., AMNH 6341*

This specimen is the most complete individual probably referable to Barosaurus (McIntosh, 2005). It was collected in three parts and subsequently separated among three institutions (USNM, CM, and UUVP), but later brought together by B Brown for the AMNH (Bird, 1985). Some doubts exist concerning the correct attribution of a tibia-fibula pair, which might belong to a Diplodocus specimen found in the vicinity of AMNH 6341 (McIntosh, 2005).

Barosaurus sp., AMNH 7530*

Both the holotype specimen of Kaatedocus siberi (SMA 0004) and AMNH 7530 were found at Howe Quarry (Michelis, 2004; Tschopp & Mateus, 2013b). AMNH 7530 is tagged as cf. Barosaurus on display at AMNH, probably based on a tentative identification made by Brown (1935), but without detailed study. Furthermore, the current display label wrongly identifies the specimens as AMNH 7535 (Michelis, 2004). AMNH 7530 is an important specimen for diplodocid taxonomy because it includes articulated anterior and mid-cervical vertebrae and a partial skull.

Barosaurus sp., AMNH 7535*

This specimen was recovered with Kaatedocus siberi SMA 0004 and AMNH 7530 at Howe Quarry (Michelis, 2004; Tschopp & Mateus, 2013b), and has been simply cataloged as Barosaurus in the collections of the AMNH (likely by B Brown; Brown, 1935). AMNH 7535 largely preserves the same elements as SMA 0004 and AMNH 7530, and appears to be of about the same size. A partial tail is also accessioned under AMNH 7535, but given the chaotic distribution of specimens in the quarry (Tschopp & Mateus, 2013a: Fig. 1), it is impossible to confidently attribute disparate and disarticulated portions to any single common individual. A diplodocid quadrate that was initially cataloged under AMNH 7535 now bears the number AMNH 30070. Because the original attribution of this quadrate to AMNH 7535 was probably based on their vicinity in the quarry, two analyses were performed with and without the information of this bone, yielding the same phylogenetic position in both iterations. In both instances, information from the caudal series was omitted from scores of AMNH 7535. Scores on the quadrate were retained in the final analysis because AMNH 30070 shows some differences with the quadrates known from Kaatedocus (e.g., lack of the small fossa dorsomedially on the quadrate shaft, E Tschopp, pers. obs., 2011), as do also the cervical vertebrae.

Barosaurus sp., CM 11984*

Together with YPM 429 and AMNH 6341, CM 11984 represents a third, relatively complete, likely Barosaurus specimen (McIntosh, 2005). Some of the material of CM 11984 is still unprepared, and further crucial information on Barosaurus can be expected once these are freed from matrix. In addition to the vertebral column, a pes is accessioned under CM 11984, which McIntosh (2005) considered to have a dubious association with the remaining material, given the chaotic quarry situation at Dinosaur National Monument. Therefore, this pes is not considered as part of the scoring of CM 11984.

Barosaurus sp., SMA O25-8*

This specimen is a partial skull from the Howe Quarry. Due to differences both in braincase and endocast morphology compared to the holotype of Kaatedocus siberi SMA 0004, Schmitt et al. (2013) showed that two diplodocine taxa were present at the Howe Quarry. SMA O25-8 was tentatively referred to Barosaurus because the elongate cervical vertebrae of the specimen AMNH 7535 (which is different from K. siberi, see above) are more similar to this genus than to any other North American diplodocine (Schmitt et al., 2013).

Brachiosaurus sp., SMA 0009*

Initially described as a diplodocid (Schwarz et al., 2007), a reassessment of the systematic position of SMA 0009 after further preparation of the mid-cervical vertebrae revealed probable titanosauriform affinities (Carballido et al., 2012a). Carballido et al. (2012a) suggested that SMA 0009 represents an immature Brachiosaurus. Therefore, B. altithorax (Riggs, 1904; Taylor, 2009) was included in our dataset to test this possibility.

Brontosaurus amplus, YPM 1981*

The type of B. amplus (Marsh, 1881) is generally referred to Apatosaurus excelsus (Gilmore, 1936; McIntosh, 1990a; McIntosh, 1995; Upchurch, Tomida & Barrett, 2004), but has never been described in detail.

Brontosaurus excelsus, YPM 1980*

The holotype of Brontosaurus excelsus (now commonly synonymized with Apatosaurus) was the first to be published with a reconstruction of the entire skeleton (Marsh, 1883) and is still one of the best preserved diplodocid specimens worldwide. The skeleton was extensively reconstructed prior to being mounted at the YPM. Therefore, special care was taken when scoring characters from the original specimen.

Camarasaurus grandis, YPM 1901

Marsh (1877a) initially assigned this species to Apatosaurus, but subsequently referred it to Morosaurus (Marsh, 1878; later synonymized with Camarasaurus: Mook, 1914). There is some confusion about the correct assignment of several bones to either the holotype YPM 1901 or the referred specimens YPM 1902 or YPM 1905 from the same quarry (see Ostrom & McIntosh, 1966). Herein, scores are included from all elements potentially belonging to YPM 1901 (according to Ostrom & McIntosh, 1966). Because all three specimens were referred to Camarasaurus, this should have no influence on the ingroup relationships of the current phylogenetic analysis.

Cetiosauriscus stewarti, NHMUK R3078*

The holotype specimen was first described in the early 1900s (Woodward, 1905) as Cetiosaurus leedsi. However, Huene (1927) identified ‘Cetiosaurus’ leedsi as a separate genus, Cetiosauriscus, and highlighted the then referred specimen NHMUK R3078 as exemplifying the new genus. NHMUK R3078 was made the holotype of Cetiosauriscus stewarti (Charig, 1980), which later was instated as the type species of Cetiosauriscus (Charig, 1993). It was included in Diplodocidae by McIntosh (1990b), based on pedal morphology, but subsequent analyses proposed a closer relationship with the non-neosauropod eusauropods Mamenchisaurus or Omeisaurus, as well as with Tehuelchesaurus (Heathcote & Upchurch, 2003). Mamenchisaurus and Omeisaurus were thus included in the present analysis in order to test these competing hypotheses. A detailed restudy of C. stewarti is in preparation by P Upchurch, P Mannion & J Heathcote (pers. comm., 2011, 2012), and will doubtlessly reveal more valid comparisons. Because personal observation of the caudal vertebrae of Spinophorosaurus nigerensis revealed high similarity with Cetiosauriscus, S. nigerensis was added to the matrix, in order to appraise the phylogenetic significance of their morphological similarities.

Dinheirosaurus lourinhanensis, ML 414*

The holotype of Dinheirosaurus lourinhanensis was originally referred to Lourinhasaurus alenquerensis by Dantas et al. (1998), but Bonaparte & Mateus (1999) realized that ML 414 represents a different genus. Contrary to the phylogenetic assignment of L. alenquerensis, which is now thought to be a basal macronarian (see below), the diplodocid affinities of D. lourinhanensis are well supported by four phylogenetic analyses (Rauhut et al., 2005; Whitlock, 2011a; Mannion et al., 2012; Tschopp & Mateus, 2013b).

Diplodocinae indet., SMA 0011*

SMA 0011 has been mentioned by Klein & Sander (2008) as Diplodocinae indet, and its ontogenetic stage identified histologically as HOS 9, corresponding to sexual maturity (Klein & Sander, 2008). The specimen is nearly complete and largely articulated, preserving bones from all skeletal regions except for the tail (E Tschopp, pers. obs., 2011). It thus plays a very important role in increasing character overlap between the more fragmentary OTUs.

Diplodocinae indet., SMA 0087*

This specimen comprises a completely articulated skeleton from mid-dorsal vertebrae to mid-caudal vertebrae, the pelvic girdle and left hindlimb. It was found at the Howe-Scott quarry, about one meter below the specimen SMA 0011 (E Tschopp, pers. obs., 2003). The histology of SMA 0087 was studied by Klein & Sander (2008), who showed that it was an adult individual (HOS 11), and identified it as Diplodocinae indet.

Diplodocus carnegii, CM 84*

The holotype of D. carnegii is one of a few specimens of Diplodocus that includes cervical vertebrae. It is mounted at CM, and has been “completed” with bones from various other specimens: CM 94, 307, 21775, 33985, HMNS 175, USNM 2673, and AMNH 965 (McIntosh, 1981; Curtice, 1996). Scores of the holotype of D. carnegii are based on this mounted specimen, with effort taken to ensure that only material from CM 84 was included. D. carnegii was erected based on comparisons to AMNH 223, which showed some differences in caudal neural spine orientation. If compared with the original type material, the differences are not as clear, and were in fact disputed by Gilmore (1932).

Diplodocus carnegii, CM 94*

This specimen was described as a cotype of D. carnegii by Hatcher (1901). Both holotype and cotype specimens were found in the same quarry, alongside material of other genera (Hatcher, 1901). Oddly, CM 94 includes two pairs of ischia, which casts some doubt on the true attribution of bones to individual specimens (McIntosh, 1981; E Tschopp, pers. obs., 2011). Because both pairs of ischia show the same characteristics, we included the entire material excluding one pair of ischia from the OTU representing CM 94 (including some bones mounted with the holotype of ‘Diplodocus’ hayi HMNS 175, see below). However, further studies are needed in order to definitively assign the various bones among the at-least two individuals present.

Diplodocus cf. carnegii, WDC-FS001A*

This specimen has not been described entirely, but is the most complete specimen referred to Diplodocus that has a manus with associated hindlimb and axial material (Bedell & Trexler, 2005). The specimen was found in two spatial clusters in the quarry, but the lack of duplicated bones, the two similarly sized humeri, and osteological indications of a single ontogenetic stage led Bedell & Trexler (2005) to identify the materials as belonging to a single individual with affinities to D. carnegii.

“Diplodocus” hayi, HMNS 175*

The holotype specimen of ‘D.’ hayi was initially housed at CM (as CM 662), prior to residing in Cleveland for a time (formerly CMNH 10670). Holland (1924) described it as a novel species of Diplodocus, based solely on cranial characters. At that time, Apatosaurus was thought to have a Camarasaurus-like skull (see Berman & McIntosh, 1978), which probably influenced researchers to identify any elongate, diplodocid skull as Diplodocus. McIntosh (1990a), amongst others, later suggested that ‘D.’ hayi might actually not belong to Diplodocus, but to a unique genus, based on various similarities with Apatosaurus in the cranium, forelimb, and tail. Because the specimen is mounted at HMNS (together with reconstructions and original bones from CM 94; McIntosh, 1981), it is only of limited accessibility. Nevertheless, the present phylogenetic analysis corroborates a referral of ‘D.’ hayi to a unique genus (see below).

Diplodocus lacustris, YPM 1922*

The original type material of D. lacustris comprises teeth, a premaxilla, and a maxilla (Marsh, 1884). However, personal observations at YPM reveal that the cranial bones clearly belong to Camarasaurus or a morphologically similar taxon, and that there is no relationship between them and the teeth. Mossbrucker & Bakker (2013) described a newly found putative apatosaur maxilla and two premaxillae from the same quarry, proposing that they might belong to the same individual as the teeth of YPM 1922. However, given the lacking field notes from the first excavations, such a referral will be difficult to prove. Therefore, in the present analysis, only the teeth were scored for D. lacustris.

Diplodocus longus, YPM 1920*

YPM 1920 constitutes the genoholotype of Diplodocus (Marsh, 1878; i.e., D. longus is the type species of Diplodocus) and thus has special taxonomic importance. Unfortunately, it is highly incomplete, with only two nearly complete caudal vertebrae, and few additional fragmentary anterior to mid-caudal vertebrae identifiable in the YPM collections. A chevron was reported as belonging to the same individual (Marsh, 1878; McIntosh & Carpenter, 1998), but it could not be located at YPM in 2011. Other articulated vertebrae were found in the field but discarded due to their friable preservation (McIntosh & Carpenter, 1998). Extraneous materials were once assigned to the same specimen, including a skull, femur, tibia, fibula, astragalus, and five metatarsals (still accessioned under YPM 1920), as well as an ulna, radius, and partial manus assigned to YPM 1906 (McIntosh & Carpenter, 1998). However, only the caudal series and the chevron can be confidently identified as belonging to the holotypic individual (McIntosh & Carpenter, 1998), as scored in the present analysis.

Diplodocus sp., AMNH 223*

This specimen was first described as Diplodocus longus (Osborn, 1899). It was the first reasonably articulated specimen of Diplodocus and thus became an important comparative specimen (see Hatcher, 1901). Three partial cervical neural arches, described and figured by Osborn (1899), were not located at AMNH during the collection visits in 2010 and 2011. Coding of these elements is thus based entirely on Osborn (1899).

Diplodocus sp., AMNH 969*

This skull and associated atlas and axis were identified as D. longus, based on an earlier report of a skull allegedly belonging to the holotype specimen of D. longus, YPM 1920 (Marsh, 1884; Holland, 1906). However, the only reported Diplodocus specimen with an articulated skull and anterior cervical vertebrae is CM 3452, of which only the skull has been described (Holland, 1924). Because no anterior cervical vertebrae are definitely attributable to D. longus, the only comparison that can be made is with the D. carnegii type specimens, of which only CM 84 preserves the axis. Because the two differ in morphology (e.g., of the prespinal lamina), AMNH 969 was herein regarded Diplodocus sp.

Diplodocus sp., CM 3452*

On display at CM, this specimen is the only possible Diplodocus with articulated skull and anterior cervical vertebrae (McIntosh & Berman, 1975). However, the cervical vertebrae have not been described, and no detailed study has been done in order to identify the species affinity for CM 3452. Comparison with other specimens referred to Diplodocus is hampered due to the presence of very little anatomical overlap.

Diplodocus sp., CM 11161*

This specimen is only a skull. It was described as Diplodocus longus by Holland (1924) and McIntosh & Berman (1975), based on comparisons with the earlier reported putative Diplodocus skulls AMNH 969, USNM 2672, and 2673. However, because all of them were disarticulated and found in quarries that also produced other diplodocid genera, care must be taken concerning these identifications. Our knowledge of diplodocid skulls to date suggests that they are extremely similar to each other, and very few distinguishing characters have yet been proposed (Berman & McIntosh, 1978; McIntosh, 2005; Harris, 2006a; Remes, 2006; Whitlock, Wilson & Lamanna, 2010; Whitlock, 2011b; Tschopp & Mateus, 2013b; Whitlock & Lamanna, 2012). Thus, we refrain from referring CM 11161 to any species of Diplodocus until postcranial diagnostic traits are robustly linked to cranial morphologies.

Diplodocus sp., CM 11255*

This skull was found without associated postcranial material, in the same quarry as the skulls CM 11161 and 11162. It was first mentioned and figured by Holland (1924), and completely described by Whitlock, Wilson & Lamanna (2010). The latter authors identified CM 11255 as Diplodocus due to obvious differences with skulls referred to Apatosaurus, Suuwassea, and Tornieria, and closer resemblance to skulls referred to Diplodocus (Whitlock, Wilson & Lamanna, 2010). However, Whitlock, Wilson & Lamanna (2010) also acknowledged that several diplodocine taxa are not known from cranial material, so that a definitive assignment to the genus Diplodocus is currently impossible.

Diplodocus sp., DMNS 1494*

This specimen is a relatively complete, articulated find from the Dinosaur National Monument. The only disarticulated elements are the right scapulacoracoid and the left hindlimb. These elements were not included in the present analysis because DMNS 1494 was found intermingled with other skeletons (V Tidwell, pers. comm., 2010). DMNS 1494 was collected by the Carnegie Museum and later transferred to DMNS for exhibit. A right fibula and astragalus of the same specimen remained at CM (presently CM 21763; McIntosh, 1981). The specimen has never been formally described, but is ascribed to D. longus (e.g., Gillette, 1991). Together with CM 84, DMNS 1494 is the only Diplodocus specimen included here with articulated, and complete cervical vertebrae.

Diplodocus sp., USNM 2672*

Like AMNH 969, USNM 2672 preserves a partial skull and atlas. It was the first diplodocid skull to be reported, and was initially included within the holotype of D. longus, YPM 1920 (Marsh, 1884), although labeled YPM 1921 (Berman & McIntosh, 1978). However, this skull and the holotypic caudal vertebrae were not found in articulation or even close association, so this attribution must be regarded as questionable (McIntosh & Carpenter, 1998), and the two specimens were treated as distinct OTUs in our analyses.

Diplodocus sp., USNM 2673*

This specimen was found in the same quarry as USNM 2672, and initially cataloged as YPM 1922, before it was transferred to USNM (McIntosh & Berman, 1975). Although it bore the same YPM specimen number as the D. lacustris holotype, it cannot be from the same specimen as they were found in different quarries (Marsh, 1884; McIntosh & Berman, 1975).

Diplodocus sp., USNM 10865*

Although USNM 10865 is one of the most complete Diplodocus specimens, it has only been preliminarily described and was tentatively referred to D. longus by Gilmore (1932). USNM 10865 was found close to the articulated Barosaurus AMNH 6341 (‘#340’ in Gilmore, 1932; McIntosh, 2005). According to McIntosh (2005), two sets of left lower legs of different lengths were found associated with USNM 10865. The shorter set was mounted by Gilmore (1932), but McIntosh (2005) suggests that this assignment might have been wrong. For our character 440 relating to the tibia/femur length, the higher ratio was therefore used, following McIntosh (2005).

Dyslocosaurus polyonychius, AC 663*

The only specimen of this putative diplodocid sauropod consists solely of appendicular elements of dubious origin and association (McIntosh, Coombs & Russell, 1992). No field notes exist, but personal observations of differing color and preservation among individual bones led to the conclusion that at least the supposed php III-1 was probably not collected at the same place as the rest of the holotype specimen (E Tschopp, 2011, unpublished data). It is therefore excluded from scores of Dyslocosaurus in this phylogenetic analysis. A more detailed reassessment of this specimen is in progress (E Tschopp & J Nair, 2015, unpublished data), and might reveal additional information on its taxonomic affinities. The phylogenetic position yielded in the present analysis is regarded as preliminary.

Dystrophaeus viaemalae, USNM 2364*

This specimen is highly fragmentary, but was identified as possibly diplodocoid by McIntosh (1990b; his ‘Diplodocidae’ conforms to the current use of the Diplodocoidea). The type material is only partly prepared, which largely impedes the identification of crucial character states. The type locality was relocated in the mid-1990s, and more material of the probable holotypic individual was excavated, of which only a phalanx has been identifiable (Gillette, 1996a; Gillette, 1996b). However, Gillette (1996a) and Gillette (1996b) stated that more material is probably present, such that additional information on Dystrophaeus might be forthcoming. Both in the initial description (Cope, 1877b) and a reassessment (Huene, 1904), several of the bones were misidentified: metacarpal V (according to Huene, 1904) is most probably a metacarpal I, based on the angled distal articular surface (McIntosh, 1997; E Tschopp, pers. obs., 2011). Cope (1877b) correctly identified a partial scapula (contra Huene, 1904, who thought it was a pubis), but misidentified a complete ulna and a partial radius as humerus and ulna, respectively, as already recognized by Huene (1904). The OTU as included here therefore consists of a partial dorsal vertebra, a partial scapula, an ulna, a distal radius, and the metacarpals.

Dystylosaurus edwini, BYU 4503*

The holotype of Dystylosaurus edwini is an anterior dorsal vertebra (Jensen, 1985). There is some doubt concerning its taxonomic affinities: it has been identified as either brachiosaurid (Paul, 1988; McIntosh, 1990b; Upchurch, Barrett & Dodson, 2004; Chure et al., 2006) or diplodocid, possibly even from the same individual as the Supersaurus vivianae holotype scapulacoracoid (Curtice & Stadtman, 2001; Lovelace, Hartman & Wahl, 2007). It was included in a preliminary analysis as an OTU independent from Supersaurus vivianae BYU and WDC DMJ-021 in order to clarify its taxonomic status. The results yielded 102 most parsimonious trees, where Dystylosaurus always grouped with the two Supersaurus OTUs, which sometimes included Dinheirosaurus ML 414, “Diplodocus” hayi HMNS 175, Barosaurus affinis YPM 419, or Diplodocus lacustris YPM 1922 within the same branch. In 31 out of 102 most parsimonious trees Dystylosaurus and the two Supersaurus OTUs were found as sister taxa. This result corroborates the hypothesis of Curtice & Stadtman (2001) and Lovelace, Hartman & Wahl (2007) that the Dystylosaurus holotypic vertebra is Supersaurus, and most probably from the same individual as the Supersaurus holotype. In our definitive analysis, BYU 4503 was thus included as part of the combined OTU representing the BYU specimens of Supersaurus vivianae.

“Elosaurus” parvus, CM 566*

CM 566 is a small juvenile that is generally referred to Apatosaurus excelsus (McIntosh, 1995), or constitutes the independent species Apatosaurus parvus together with an adult specimen (UW 15556; Upchurch, Tomida & Barrett, 2004), with which it was found associated (Peterson & Gilmore, 1902). However, it was initially described as a unique genus (Peterson & Gilmore, 1902).

Eobrontosaurus yahnahpin, Tate-001

Initially described as Apatosaurus yahnahpin (Filla & Redman, 1994), a separate genus was erected for the specimen (Bakker, 1998), partly based on differences in coracoid morphology to Apatosaurus. The specimen has been considered a camarasaurid (Upchurch, Barrett & Dodson, 2004), but more recently, Mannion (2010) suggested diplodocid affinities. The taxon has never been included in any phylogenetic analysis, but a detailed description of the entire material appears to be in preparation (R Bakker, pers. comm., 2008, cited in Mannion, 2010).

Kaatedocus siberi, SMA 0004*

Before its detailed examination, the holotype of Kaatedocus siberi was generally reported as Diplodocus (Ayer, 2000) or Barosaurus (Michelis, 2004). Subsequently, a description and phylogenetic reappraisal of SMA 0004 revealed its generic separation from Diplodocus and Barosaurus (Tschopp & Mateus, 2013b).

Kaatedocus siberi, SMA D16-3*

This additional specimen from the Howe Quarry (a partial skull) was referred to K. siberi by Schmitt et al. (2013). The skull bones were found disarticulated but associated (E Tschopp, pers. obs., 2012), and have not been described in detail yet.

Leinkupal laticauda, MMCH-Pv 63-1

The holotype of Leinkupal laticauda was only recently described (Gallina et al., 2014). It includes only a single caudal vertebrae, although more elements from the same quarry were referred to the species by Gallina et al. (2014). All diplodocid remains were found disarticulated and mingled with dicraeosaur material (Gallina et al., 2014), and it is thus currently too early to include more than the holotypic anterior caudal vertebra in a specimen-level cladistic analysis as attempted herein.

Losillasaurus giganteus, MCNV Lo-1 to 26*

This OTU represents an individual containing the holotypic caudal vertebra, Lo-5, the paratypes Lo-10 and Lo-23, and several additional elements. All the bones of MCNV Lo-1 to 26 were found associated and no duplication of bones occurred (Casanovas, Santafé & Sanz, 2001). Initially regarded as a basal diplodocoid (Casanovas, Santafé & Sanz, 2001), Losillasaurus was soon found to represent a non-diplodocoid, and probably a non-neosauropod eusauropod (Rauhut et al., 2005; Harris, 2006c). With the description of Turiasaurus (Royo-Torres, Cobos & Alcalá, 2006), which has since been consistently recovered as sister genus to Losillasaurus (Royo-Torres, Cobos & Alcalá, 2006; Royo-Torres et al., 2009; Barco, 2009; Carballido et al., 2012b; Royo-Torres & Upchurch, 2012), this more basal position has been generally accepted. Therefore, Turiasaurus was added as an outgroup to test their sister relationship.

Lourinhasaurus alenquerensis, lectotype*

This species was first described by Lapparent & Zbyszewski (1957) as referable to Apatosaurus, but later included in Camarasaurus (McIntosh, 1990a). Subsequently, Dantas et al. (1998) erected a new genus for the species, but only Antunes & Mateus (2003) clearly assigned a specific type specimen to the species. Lourinhasaurus has usually been recovered as a basal macronarian in recent phylogenetic analyses (Royo-Torres & Upchurch, 2012; Mocho, Royo-Torres & Ortega, 2014).

“Seismosaurus” hallorum, NMMNH 3690

The holotype of S. hallorum was initially described as S. halli, and as one of the largest sauropods ever (Gillette, 1991). However, this identification as a unique genus, and its size estimate, were mainly based on an incorrect assignment of the position of some mid-caudal vertebrae (Curtice, 1996; Herne & Lucas, 2006). Subsequent reanalysis of the specimen revealed that it is indistinguishable from Diplodocus and that it probably belongs to the same species as AMNH 223 and USNM 10865 (Lucas et al., 2006; Lovelace, Hartman & Wahl, 2007). Gillette himself (1994) corrected the species name from halli to hallorum, as he did not apply the correct latin ending for the plural in the initial description (Gillette, 1991; Gillette, 1994). Because the corrected name has since been used more widely than the original proposal, it is followed here. Herne & Lucas (2006) added a femur (NMMNH 25079) from the same quarry to the holotype individual, which is also used to score the taxon in the analysis herein.

Supersaurus vivianae, BYU (various specimen numbers)*

Supersaurus vivianae is based on a scapulacoracoid (Jensen, 1985; Curtice, Stadtman & Curtice, 1996; Curtice & Stadtman, 2001; Lovelace, Hartman & Wahl, 2007). It was found at the Dry Mesa Quarry, intermingled with other large bones of diplodocid, brachiosaurid, and camarasaurid affinities (Jensen, 1985; Jensen, 1987; Jensen, 1988; Curtice & Stadtman, 2001). Jensen (1985) described three new taxa based on this material: Supersaurus vivianae, Dystylosaurus edwini, and Ultrasauros macintoshi. Subsequent study of the Dry Mesa specimens indicates that the holotypic dorsal vertebra of Dystylosaurus, as well as a dorsal vertebra referred to Ultrasauros by Jensen (1985) and Jensen (1987) probably belonged to the same individual as the holotypic scapulacoracoid of Supersaurus vivianae (Curtice & Stadtman, 2001). Lovelace, Hartman & Wahl (2007) revised this referral based on a new find from Wyoming, agreeing in large parts with Curtice & Stadtman (2001). The revised composition of the holotypic individual is listed in the Table 4. Since a preliminary analysis of the phylogenetic affinities of Dystylosaurus (see above) further corroborated this referral, a combined OTU was used for the final analysis.

Supersaurus vivianae, WDC DMJ-021*

WDC DMJ-021 is a reasonably articulated skeleton and represents the most complete specimen of S. vivianae (Lovelace, Hartman & Wahl, 2007). It is not directly comparable with the holotype, because no scapulacoracoid was found. Nevertheless, based on the overlap with additional material attributed to the holotypic individual (see above; Lovelace, Hartman & Wahl, 2007), the identification of WDC DMJ-021 as S. vivianae has been widely accepted.

Suuwassea emilieae, ANS 21122*

Suuwassea was initially identified as a flagellicaudatan with uncertain affinities to Diplodocidae or Dicraeosauridae (Harris & Dodson, 2004). Further analyses suggest a closer relationship with the Dicraeosauridae (Salgado, Carvalho & Garrido, 2006; Whitlock & Harris, 2010; Whitlock, 2011a), which would mean that Suuwassea is the only North American representative of this taxon.

Tornieria africana, holotype (various specimen numbers)*

The holotype specimen of T. africana was found at the locality “A” at Tendaguru, Tanzania (Fraas, 1908; Remes, 2006). Tornieria was initially described as Gigantosaurus africanus (Fraas, 1908), but Sternfeld (1911) noted that this generic name was preoccupied, proposing the combination T. africana as a replacement. Janensch (1922) suggested synonymy of Tornieria and Barosaurus, resulting in the combination Barosaurus africanus, and later referred much more material from various quarries to the same species (Janensch, 1935; Janensch, 1961). However, in a reassessment of the entire material, which also resurrected the name Tornieria africana, only two or three individuals were positively identified as belonging to Tornieria (Remes, 2006). Remes (2006) furthermore identified additional material from the same quarry as most probably belonging to the same individual as the holotype. We therefore follow Remes (2006) by including all the Tornieria material found at locality “A” in the holotypic OTU (Table 4).

Tornieria africana, skeleton k*

A second specimen of T. africana comes from the “k” quarry at Tendaguru and was the only individual found at that site (Heinrich, 1999; Remes, 2006). Initially relatively complete with semi-articulated vertebral column and numerous appendicular elements, much of it has been lost or was destroyed during World War II (Remes, 2006). For these elements, descriptions and figures in Janensch (1929b) were used to complement the scoring.

Skull

C1: Premaxillary anterior margin, shape: without step (0); with marked but short step (1); with marked and long step (2) (U98-10; W98-19; modified by C12b-2; Fig. 1). Ordered.

Comments. The character describes the presence and development of a horizontal portion of the premaxilla, which lies anterior to the nasal process. The step, when present, is best visible in lateral view. It was initially proposed by Upchurch (1998), who scored the Diplodocoidea as unknown or inapplicable, due to a supposed absence of the nasal process. However, some diplodocoids, (e.g., Suuwassea) clearly show a distinction between the anterior main body and the posterior nasal process in dorsal view, where they show an abrupt narrowing (Harris, 2006a; ANS 21122, E Tschopp, pers. obs., 2011). Diplodocoidea should therefore be scored as ‘0.’ A third state was added in order to distinguish Brachiosauridae from other macronarian sauropods (Carballido et al., 2012b). The character is treated as ordered, due to the gradational change in morphology.

C2: Premaxilla, external surface: without anteroventrally orientated vascular grooves originating from an opening in the maxillary contact (0); vascular grooves present (1) (Wilson, 2002; S07-3; Fig. 2).

Comments. The presence of these grooves was previously found as a synapomorphy of Dicraeosauridae (Whitlock, 2011a; Mannion et al., 2012). However, faint grooves originating at the premaxillary-maxillary contact are also visible in Nigersaurus (Sereno et al., 2007) and in some diplodocid specimens. In the latter, they fade shortly anterior to the suture (e.g., in CM 11161, 11162, SMA 0011, USNM 2672). In the present analysis, all of these specimens are scored as apomorphic.

C3: Premaxilla, shape in dorsal view: main body massive, with proportionally short ascending process distinct (0); single elongate unit, distinction between body and process nearly absent (1) (U98-12; wording modified; Fig. 2).

Comments. Upchurch (1998) formulated this character differently, based on his interpretation that the ascending process of the premaxilla was absent in Diplodocoidea. As stated above, this is not the case. The wording of the derived state was thus changed accordingly.

C4: Premaxilla, angle between lateral and medial margins of premaxilla as seen in dorsal view: >40° (0); 17° −40° (1); <17° (2) (Upchurch, 1999; modified; Table S2). Unordered.

Comments. Upchurch (1999) was the first to note significant differences in these angles between diplodocoids (around 10°), nemegtosaurids (18°), and remaining taxa (e.g., Giraffatitan, 30°; Upchurch, 1999: Fig. 7). He used this character (with two states) as one of several that supported the inclusion of Nemegtosauridae within Diplodocoidea (Upchurch, 1999), a view now falsified by nearly complete finds of new nemegtosaurids that show them to be deeply nested within titanosaurians, but with convergences with Diplodocoidea (Wilson, 2002; Curry Rogers, 2005; Zaher et al., 2011). The OTUs included in this dataset were rescored for this character based on figures or on original material. Because the lateral margin is concave to sinuous in most taxa, a straight line was drawn from the anterior-most point of the premaxillary-maxillary contact to the point where the lateral edge curves medially, at the base of the ascending process. The results (Table S2) indicate that the distribution of the character scores is not as straightforward as previously thought: Shunosaurus, as well as some specimens of Camarasaurus appear to show similarly narrow angles as Dicraeosaurus and Suuwassea. A third state was thus added, such that diplodocid and rebbachisaurid OTUs now score in the narrow-most range, and Mamenchisaurus and Jobaria are classed as significantly wide-angled taxa. Because the derived state is ambiguous, the character is most parsimoniously left unordered.

C5: Premaxilla, posteroventral edge of ascending process in lateral view: concave (0); straight and dorsally oriented (1); straight, and directed posterodorsally (2) (W11-3; wording modified; Fig. 1). Unordered.

Comments. Whitlock (2011a: p.35) described the character as follows: ‘Ascending process of the premaxilla, shape in lateral view: convex (0); concave, with a large dorsal projection (0); sub-rectilinear and directed posterodorsally (1).’ This formulation is misleading, and the states overlap with those of character 1, which describes the premaxillary ‘step.’ Varying morphologies of the ascending process, following the states of Whitlock (2011a), were observed among the included taxa regarding the posteroventral edge of the ascending process—the margin that delimits the nasal opening anteriorly. The description of the character was adapted, reducing the character to only encompass the orientation of the posteroventral edge, thereby avoiding overlap with character 1. The directional terms in the states are meant in relation to a horizontally oriented ventral edge of the maxilla. Because no state is obviously intermediate relative to the other two, the character is left unordered.

C6: Premaxilla, posterolateral process and the lateral process of the maxillary, shape: without midline contact (0); with midline contact forming a marked narial depression, subnarial foramen not visible laterally (1) (W02-1; Fig. 3).

Comments. Whitlock (2011a) reversed the polarity of this character, due to a more limited outgroup sampling. With the inclusion of Shunosaurus (Mannion et al., 2012), the most basal OTU again lacks the midline contact, as is the case in Diplodocoidea. The original phrasing of Wilson (2002) is therefore preferred.

C7: Premaxilla, dorsoventral depth of anterior portion: remains the same as posteriorly, or widens gradually (0); widens considerably, and abruptly (1) (Harris, 2006a; Fig. 4).

Comments. Harris (2006a) stated this difference as useful to distinguish Suuwassea (which retains the same depth) from Diplodocus (which widens). A similar, narrow premaxilla is furthermore present in Kaatedocus (Tschopp & Mateus, 2013b). The character is difficult to observe in articulated skulls, but single elements do show a significant difference.

C8: Subnarial foramen and anterior maxillary foramen, position: well distanced from one another (0); separated by narrow bony isthmus (1) (W02-5; Fig. 5).

C9: Maxilla, large foramen anterior to the preantorbital fossa, separated by a narrow bony bridge: absent (0); present (1) (Z11-244; wording modified; Fig. 3).

Comments. Generally, sauropod maxillae are pierced by a number of small foramina anteriorly, probably for innervation and/or blood supply of the replacement teeth. The foramen described by Zaher et al. (2011) in Tapuiasaurus, however, is relatively large, and closely attached to the preantorbital fossa. The same is the case in Dicraeosaurus hansemanni MB.R.2336 (Janensch, 1935), but not in diplodocids.

C10: Maxilla, large foramen posterior to anterior maxillary foramen, dorsal to preantorbital fossa: absent (0); present (1) (New; Fig. 3).

Comments. Few diplodocid specimens show a large foramen posterior to the anterior maxillary foramen (e.g., Kaatedocus SMA 0004). This foramen cannot be the same as the one described in character 9, given that both are present in Dicraeosaurus.

C11: Anterior maxillary foramen, location: detached from maxillary-premaxillary boundary, facing dorsally (0); lies on medial edge of maxilla, opening medially into the premaxillary-maxillary boundary (1) (New; Fig. 3).

Comments. Usually, diplodocids have the subnarial and the anterior maxillary foramina enclosed within a single, elongated fossa at the maxillary-premaxillary boundary (Wilson & Sereno, 1998; Whitlock, 2011b). However, in Kaatedocus, the anterior maxillary foramen is detached and laterally positioned, within a unique, small fossa. It thus resembles the plesiomorphic state present in Jobaria or Camarasaurus (Wilson & Sereno, 1998; Sereno et al., 1999), although it is still much closer to the subnarial foramen. Primitive outgroup taxa (those normally basal to Jobaria) were coded as unknown, as it is unclear if the intermaxillary foramen that is present in these taxa (e.g., He, Li & Cai, 1988; Ouyang & Ye, 2002) is homologous to the anterior maxillary foramen or the subnarial foramen.

C12: Maxilla, canal connecting the antorbital fenestra and the preantorbital fossa: absent (0); present (1) (New; Fig. 3).

Comments. Such a canal is only present in SMA 0011 and USNM 2673. Taxa without a preantorbital fossa were scored as unknown in order to avoid absence coding.

C13: Maxilla, dorsal process, posterior extent: anterior to or even with posterior process (0); extending posterior to posterior process (1) (W11-9; Fig. 1).

Comments. The character is applied to skulls in lateral view, with the ventral edge of the maxilla oriented horizontally.

C14: Maxilla-quadratojugal contact: absent or small (0); broad (1) (Y93-14; Fig. 1).

Comments. Upchurch (1998) reported some difficulties in scoring some taxa for his version of this character, which was defined as a simple absence-presence feature. Reduced, small contacts are present in Camarasaurus, but only diplodocids are known to have developed a broad area where the maxilla contacts the quadratojugal (Upchurch, 1998; Wilson & Sereno, 1998). Therefore, Whitlock (2011a) redefined the states, such that the apomorphic state now describes a synapomorphy of at least Diplodocidae (it is unknown in Dicraeosauridae and Rebbachisauridae). The derived state appears to be a convergence in some nemegtosaurids (Upchurch, 1998; Wilson, 2005).

C15: Preantorbital fossa: absent (0); present (1) (T13-10; Fig. 1).

Comments. Although some flagellicaudatan taxa have reduced to entirely closed preantorbital fenestrae, all show a distinct fossa, which is otherwise only present in some nemegtosaurids (Wilson, 2005).

C16: Preantorbital fossa, if present: with relatively indistinct borders (0); dorsally capped by a thin, distinct crest (1) (Wilson, 2002; W11-12; modified; Fig. 3).

Comments. Wilson (2002) originally proposed that the presence of a dorsally capped preantorbital fenestra is an autapomorphy of Diplodocus. A broader survey of this character shows that within Flagellicaudata, the absence of this dorsal crest is instead only known from a single Apatosaurus skull (CM 11162), and thus might represent an autapomorphy of Apatosaurus louisae.

C17: Preantorbital fenestra: reduced to absent (0); present, occupying at least 50% of the preantorbital fossa (1) (Berman & McIntosh, 1978; Y93-21; modified; Fig. 3).

Comments. Yu (1993) was the first to use this feature in a phylogenetic analysis. Tschopp & Mateus (2013b) modified the character, and included the dorsal crest as well. However, because these two features are not correlated (Kaatedocus has a dorsal crest but a reduced to absent fenestra), the states were adjusted, and a ratio is given to distinguish the small opening in Dicraeosaurus from the large ones in Diplodocus, for example. Large preantorbital fenestrae are convergently present in nemegtosaurids (Wilson, 2005; Zaher et al., 2011).

C18: Antorbital fenestra, maximum diameter: much shorter (<90%) than orbital maximum diameter (0); subequal (>90%) to orbital maximum diameter (1) (Y93-7; modified; Table S3).

Comments. Yu (1993) proposed the character without any clear state boundaries, which were later added by Whitlock (2011a), and changed herein from 85% to 90% in order to include Mamenchisaurus within the plesiomorphic state.

C19: Antorbital fenestra, anterior extension: is restricted posterior to preantorbital fossa (0); reaches dorsal to preantorbital fossa (1) (New; Fig. 1).

Comments. The character has to be scored with the ventral border of the maxilla oriented horizontally. Within flagellicaudatans, the derived state is most developed in Kaatedocus SMA 0004, but nemegtosaurids like Rapetosaurus have extremely elongated antorbital fenestrae that even reach anterior to the entire preantorbital fossa (Curry Rogers & Forster, 2004).

C20: Antorbital fenestra, shape of dorsal margin: straight or convex (0); concave (1) (W11-14; Fig. 1).

Comments. The diplodocine skull AMNH 969 appears to have a convex dorsal margin at first glance. However, the presence of a lateral projection in the upper half of this edge indicates that the convex shape might be due to deformation. The lateral projection in AMNH 969 is at the same location, and has the same shape as the osteological feature producing the concave dorsal edge of the antorbital fenestra in CM 11161. AMNH 969 is thus interpreted to possess the derived state, as in all flagellicaudatans.

C21: External nares, position: retracted to level of orbit, facing laterally (0); retracted to position between orbits, facing dorsally or dorsolaterally (1) (McIntosh, 1989; U95; modified by W11-15; Fig. 1).

Comments. Upchurch (1995) was the first to include this character in a phylogenetic analysis, based on observations made by McIntosh (1989). Whitlock (2011a) adjusted the state description, since the reduced taxon sampling made a third state redundant (anterior to orbit, the plesiomorphic state in Sauropoda; Upchurch, 1995).

C22: External nares, maximum diameter: shorter than orbital maximum diameter (0); longer than orbital maximum diameter (1) (U95; modified by W98-89).

Comments. Upchurch (1995) initially defined the character states in relation to skull length, but later, Wilson & Sereno (1998) changed them to relate to orbital diameter. The latter has since been widely used and is thus retained here.

C23: Prefrontal, medial margin, shape: without distinct anteromedial projection (0); curving distinctly medially anteriorly to embrace the anterolateral corner of the frontal (1) (New; Fig. 6).

Comments. In some basal sauropods, the prefrontal is located entirely anterior to the frontal. These cases are scored as plesiomorphic.

C24: Prefrontal, posterior process size: small, not projecting far posterior of frontal-nasal suture (0); elongate, approaching parietal (1) (W02-14; Fig. 7).

Comments. This character is not as straight forward as it seems. Care has to be taken that one observes the frontal and prefrontal in exactly perpendicular view. In some reconstructed dorsal views of the skull of Diplodocus (Wilson & Sereno, 1998; Whitlock, 2011b), the posterior extension of the prefrontal is remarkable, but this is due to the view in which the reconstruction is drawn. The frontal slants posteriorly, and more posterior distances therefore appear shorter. In direct dorsal view, differences in distance between taxa diminish. However, the character remains informative: in diplodocids like Apatosaurus or Diplodocus, the posterior process of the prefrontal almost reaches or surpasses the midlength of the frontal, whereas in Rebbachisauridae or in Kaatedocus and Tornieria, it remains restricted to about the anterior third (Fig. 7).

C25: Prefrontal, posterior process shape: straight (0); hooked (1) (W02-15; modified; Fig. 7).

Comments. As the posterior elongation of the prefrontal, this character was initially defined in a somewhat ambiguous way (flat/hooked). Nigersaurus does have a posteriorly facing, pointed prefrontal. The description ‘flat’ therefore does not fit very well, and it is replaced by ‘straight.’ Hooked is herein interpreted to describe a medially curving posterior process, such that its posterior end forms the medial-most extension of the prefrontal.

C26: Frontals, midline contact (symphysis): patent suture (0); fused in adult individuals (1) (Salgado & Calvo, 1992; Y93-33; Fig. 5).

Comments. Fusion of skull bones is usually considered an ontogenetic feature (Varricchio, 1997; Whitlock & Harris, 2010). However, the ontogenetic stages when fusion begins might still be different between taxa and thus phylogenetically significant. This appears to be the case here, where the braincases of Dicraeosaurus and Amargasaurus have completely obliterated sutures between the frontals, whereas large-sized diplodocid skulls do not (e.g., CM 11161). Nonetheless, it remains possible that non-dicraeosaurid sauropods fuse their frontals at an old age. In future, it might be helpful to constrict the character to a specific age-range (possibly subadult or early adult), but to date, the exact individual age of the specimens showing the fused frontals remains unknown.

C27: Frontal, anteroposterior length: long, >1.4 times minimum transverse width (0); short, 1.4 or less times minimum transverse width (1) (G86; modified; Table S4).

Comments. This character was widely used in phylogenetic analyses of sauropod dinosaurs (Upchurch, 1998; Wilson, 2002; Whitlock, 2011a; Mannion et al., 2012; Tschopp & Mateus, 2013b), with varying definitions of the state boundaries. In addition, it was often unclear if minimum or maximum transverse width was intended (e.g., Whitlock, 2011a; Tschopp & Mateus, 2013b). As shown in Table S4, there are significant differences in the ratios, with more distinct changes when comparing frontal length and minimum transverse width. Therefore, state boundaries were herein defined numerically, which also led to some differential scorings compared to Tschopp & Mateus (2013b). Kaatedocus, for example, is now well within the ratios for the apomorphic state.

C28: Frontal-nasal suture, shape: flat or slightly bowed anteriorly (0); v-shaped, pointing posteriorly (1) (W11-21; Fig. 6).

Comments. The frontals of ‘Diplodocus’ hayi might have a posteriorly pointing nasal contact as well (Holland, 1906). However, the nasals are not preserved in this specimen, and it seems thus more appropriate to score HMNS 175 as unknown.

C29: Frontals, distinct anterior notch medially between the two elements: absent (0); present (1) (T13-25; modified; Fig. 5).

Comments. The shape description of the notch (Tschopp & Mateus, 2013b) was excluded from the character in order to include also Spinophorosaurus, and SMA 0011 in the apomorphic state. The frontal usually becomes extremely thin in this part, and it is thus easily broken. Because the notch still appears genuine in these three taxa/specimens, the character was retained. Tschopp & Mateus (2013b) mentioned this feature as an autapomorphy of Kaatedocus. Given that a similar notch is present in SMA 0011, this character might actually be more widespread within Diplodocidae. In fact, many specimens (e.g., Apatosaurus CM 11162) show broken anteromedial edges in the frontal, which makes it difficult to evaluate this character. New finds of diplodocid frontals might shed some more light on the distribution of this character.

C30: Frontals, dorsal surface: without paired grooves facing anterodorsally (0); grooves present, extend on to nasal (1) (W11-22; Fig. 5).

Comments. Grooves appear to be present on the frontals of the dicraeosaurid Amargasaurus cazaui (Salgado & Calvo, 1992: Fig. 2B), but these extend onto the prefrontals and not the nasals and do not extend as far posteriorly as in Limaysaurus. Amargasaurus is thus scored as plesiomorphic, following Whitlock (2011a).

C31: Frontal, lateral edge in dorsal view: relatively straight (0); deeply concave (1) (New; Fig. 7).

Comments. This character has a somewhat ambiguous distribution. There is some difference in the shapes taken together in the plesiomorphic state as well: Rebbachisauridae, in contrast with most other taxa, have a weakly convex lateral frontal edge. Diplodocids exhibit varying shapes: Apatosaurus and Diplodocus have concave edges, whereas Kaatedocus or Tornieria have straight margins.

C32: Frontal, contribution to dorsal margin of orbit: less than 1.5 times the contribution of prefrontal (0); at least 1.5 times the contribution of prefrontal (1) (W11-23; modified by M12-20; Table S5).

Comments. The lengths of the frontal and prefrontal are measured in a straight line in lateral view, from the mid-point of the frontal-prefrontal articulation to the anterior-most (prefrontal) or posterior-most (frontal) point. Whitlock (2011a) proposed the character, leaving a gap between plesiomorphic and apomorphic states (subequal, or twice), which was changed by Mannion et al. (2012). A comparative analysis of the included specimens confirms the utility of the boundary proposed by Mannion et al. (2012).

C33: Frontal, free lateral margin: rugose (0); smooth (1) (T13-23; Fig. 7).

Comments. Rugosities are present around the dorsal margin of almost all sauropods, but in some cases, they are shifted onto the prefrontal or the postorbital. Tschopp & Mateus (2013b) hypothesized that the rugosities served for an attachment of a palpebral element.

C34: Frontal, contribution to margin of supratemporal fenestra/fossa: present (0); absent, frontal excluded from anterior margin of fenestra/fossa (1) (W98-65; Fig. 5).

Comments. In the derived state, the frontal is excluded from a contribution to the margin of the supratemporal fenestra by a contact between the medial process of the postorbital and the anterolateral process of the parietal.

C35: Frontal-parietal suture, position of medial portion: closer to anterior extension of supratemporal fenestra (0); closer to posterior extension (1) (T13-26; modified; Fig. 5).

Comments. Tschopp & Mateus (2013b) formulated the character inspired by Remes (2006), who mentioned the position of the fronto-parietal suture as a feature to distinguish Tornieria from Diplodocus. They used a tripartite character, with an intermediate state as closer to the central portion of the supratemporal fenestra (Tschopp & Mateus, 2013b). The position of the suture is difficult to assess in some diplodocid specimens, because it describes a strongly sinuous curve (e.g., CM 11161, Fig. 7). The character is thus restricted to the medial portion of the suture herein. By doing so, it becomes clear that the majority of Diplodocus skulls shifted the suture backwards, whereas all other specimens have it anteriorly located. The posterior dislocation might thus prove to be an autapomorphy of Diplodocus. The intermediate state becomes redundant, and is not included here.

C36: Pineal (parietal) foramen between frontals and parietals: present (0); absent (1) (Y93-27; modified; Fig. 5).

Comments. This character was proposed in combination with the presence of a postparietal foramen (Yu, 1993). The two are herein separated in two characters, because Kaatedocus SMA 0004 has a postparietal but no pineal foramen (Tschopp & Mateus, 2013b). The presence of a pineal foramen is often difficult to assess due to breakage of the area around the fronto-parietal suture (McIntosh, 1990b; Upchurch, Barrett & Dodson, 2004; Harris, 2006a). However, in some specimens, the presence or absence of this feature is genuine, and it thus appears appropriate to include this character. Specimens where the presence of the foramen has been doubted previously are scored as unknown. At the current state of knowledge, the presence seems to be a retained plesiomorphy characterizing the Dicraeosauridae, but in many diplodocid specimens its presence cannot be dismissed yet.

C37: Orbit, anterior-most point: anterior to the anterior extremity of lateral temporal fenestra (0); roughly even with or posterior to anterior extent of lateral temporal fenestra (1) (G86; U95; modified by W11-25; Fig. 1).

Comments. The original character was a multistate character (Upchurch, 1995). Given the limited taxon sampling of Whitlock (2011a) and the herein presented analysis, the third state becomes redundant (infratemporal fenestra restricted posterior to orbit).

C38: Orbital ventral margin, anteroposterior length: broad, with subcircular orbital margin (0); reduced, with acute orbital margin (1) (W98-25; Fig. 1).

Comments. The derived state results in a teardrop-shape of the orbit. With the ventral margin of the maxilla held horizontally, the ‘ventral margin’ would be better described with ‘anteroventral corner.’

C39: Postorbital, posterior process: present (0); absent (1) (W02-17; Fig. 1).

Comments. The postorbital is usually a triradiate bone, with a relatively short posterior process that overlaps the squamosal. The latter is absent in rebbachisaurids (Wilson, 2002; Whitlock, 2011a).

C40: Jugal, contribution to antorbital fenestra: very reduced or absent (0); large, bordering approximately one-third of its perimeter (1) (Berman & McIntosh, 1978; U95; modified by W11-28; Fig. 8).

Comments. Recognized as distinctive feature of Diplodocoidea by Berman & McIntosh (1978), the contribution of the jugal to the antorbital fenestra was first used as phylogenetic character by Upchurch (1995). Whitlock (2011a) defined the state boundaries quantitatively.

C41: Jugal, contact with ectopterygoid: present (0); absent (1) (U95; Fig. 9).

Comments. The development of this character is barely known in sauropods. When preserved, the osteology of the palatal complex is often left obscured by matrix for stability of the specimen. At the current state of knowledge, the ectopterygoid becomes anteriorly dislocated in Neosauropoda, and contacts the maxilla instead of the jugal. Future CT scanning of additional skulls will yield more detailed results.

C42: Jugal, posteroventral process: short and broad (0); narrow and elongate (1) (New; Fig. 8).

Comments. This character shows varying shapes in the skulls traditionally identified as Diplodocus (CM 11161 has a short process, whereas in all other skulls they are elongated). However, too few diplodocid jugals are preserved entirely in order to evaluate the distribution of this character to date.

C43: Jugal, dorsal process: present (0); absent (1) (Y93-24; polarity inverted; Fig. 8).

Comments. Yu (1993) proposed the dorsal process as a synapomorphy for Diplodocidae. However, no jugal is known from dicraeosaurids, and such a process is also present in Shunosaurus, Omeisaurus, and Mamenchisaurus (Janensch, 1935; He, Li & Cai, 1988; Salgado & Calvo, 1992; Chatterjee & Zheng, 2002; Ouyang & Ye, 2002). Because the latter basal taxa show dorsal processes of the jugal, the character polarity was inverted relative to the original version (Yu, 1993). Although they are scored for the plesiomorphic state, Diplodocidae is still distinguishable from Shunosaurus and the other taxa by the strong development of the dorsal process, and its anterior displacement. In Omeisaurus, e.g., the dorsal process is short and located at midlength of the jugal-lacrimal suture (He, Li & Cai, 1988).

C44: Jugal, anterior spur dorsally, which projects into antorbital fenestra: absent (0); present (1) (New; Fig. 8).

Comments. Such a spur is present in many diplodocid specimens, although in USNM 2672, it only occurs on the left side (E Tschopp, pers. obs., 2011). However, the possibility to develop such a spur still appears to be restricted to Diplodocidae, and the character is thus used in the analysis. USNM 2672 is scored as ‘present.’

C45: Quadratojugal, position of anterior terminus: anterior margin of orbit or posteriorly restricted (0); beyond anterior margin of orbit (1) (W11-30; modified; Fig. 1).

Comments. The character is coded with the ventral margin of the maxilla held horizontally. State boundaries by Whitlock (2011a: posterior to middle of orbit, anterior margin or beyond) were adjusted because all diplodocoids show strongly elongated anterior processes that end significantly anterior to the orbit. On the other hand, in Mamenchisaurus or Giraffatitan, the processes reach the anterior margin of the orbit (Janensch, 1935; Ouyang & Ye, 2002), which would require a scoring as apomorphic when following the description of Whitlock (2011a).

C46: Quadratojugal, angle between anterior and dorsal processes: less than or equal to 90°, so that the quadrate shaft is directed dorsally (0); greater than 90°, approaching 130°, so that the quadrate shaft slants posterodorsally (1) (G86; U95; Fig. 1).

Comments. The angle between the quadratojugal processes reaches its maximum in the large skulls CM 11161 and 11162. In smaller skulls (of both ontogenetically younger as well as phylogenetically more basal specimens), the angle is of approximately 110°(e.g., Kaatedocus SMA 0004; Tschopp & Mateus, 2013b), but still clearly in the derived state.

C47: Lacrimal, anterior process: absent (0); present (1) (W02-11; polarity reversed by M13-80; Fig. 1).

Comments. Wilson (2002) initially proposed the character with inverted polarity. This was changed by Mannion et al. (2013), and herein in order to have the chosen outgroups showing the plesiomorphic state. An anterior process is usually interpreted to be absent in diplodocoids. However, SMA 0011 and Dicraeosaurus do have one. On the other hand, it is possible that the feature is more widespread among Diplodocoidea, but that the anterior process is obscured by the posterodorsal process of the maxilla. The latter partly overlaps the anterior process of the lacrimal in SMA 0011. The presence of an anterior process of the lacrimal would otherwise be one of the distinguishing characteristics between diplodocoids and nemegtosaurids (Wilson, 2005).

C48: Lacrimal, dorsal portion of lateral edge: flat (0); bears dorsoventrally elongate, shallow ridge (1); bears a dorsoventrally short laterally projecting spur (2) (T13-34; Fig. 3). Ordered.

Comments. There is some evidence that this character is ontogenetically controlled (Tschopp & Mateus, 2013b): only small skulls show the laterally projecting spur. The character is retained here in order to test its validity. The character is treated as ordered due to intermediate morphologies.

C49: Quadrate, articular surface shape: quadrangular in ventral view, orientated transversely (0); roughly triangular in shape (1); thin, crescent-shaped surface with anteriorly directed medial process (2) (W11-32; Fig. 10). Ordered.

C50: Quadrate, short transverse ridge medially on posterior side of ventral ramus, close to the articular surface with the lower jaw: absent (0); present (1) (New; Fig. 11).

Comments. This ridge is a detail which appears to be synapomorphic for Diplodocidae. Most of the diplodocid quadrates could not be studied first hand for this character. Therefore a more detailed evaluation of this character has to be undertaken in order to corroborate the presence or absence of such a ridge, and its taxonomic utility.

C51: Quadrate fossa, depth: shallow (0); deeply invaginated (1) (R93-2; Fig. 11).

C52: Quadrate, shallow, second fossa medial to pterygoid flange on quadrate shaft (not the quadrate fossa): absent (0); present, becoming deeper towards its anterior end (1) (T13-37; wording modified; Fig. 12).

Comments. The medial surface of the pterygoid flange is nearly always concave, but concave dorsoventrally. In SMA 0004, as well as some other diplodocid specimens, the second fossa is transversely concave, lies anteriorly on the posterior shaft, medial to where the pterygoid flange originates. There is a chance that the character might be ontogenetic, given that no large-sized skull has yet been identified to bear this second fossa. The character was slightly reworded from its original version (Tschopp & Mateus, 2013b) in order to describe the location of the fossa better.

C53: Quadrate, dorsal margin: concave, such that pterygoid flange is distinct from quadrate shaft (0); straight, without clear distinction of posterior extension of pterygoid flange (1) (New; Fig. 12).

C54: Quadrate, posterior end (posterior to posterior-most extension of pterygoid ramus): short and robust (0); elongate and slender (1) (New; Fig. 12).

C55: Squamosal, anterior extent: restricted to postorbital region (0); extends well past posterior margin of orbit (1); extends beyond anterior margin of orbit (2) (W11-35; Fig. 1). Ordered.

Comments. The anterior extent of the squamosal is measured with the ventral border of the maxilla oriented horizontally.

C56: Squamosal-quadratojugal contact: present (0); absent (1) (U95; Fig. 13).

Comments. In diplodocids, where no contact is present, the distance between the squamosal and the quadratojugal varies (Whitlock, Wilson & Lamanna, 2010; Whitlock & Lamanna, 2012). However, most of the diplodocid specimens do not preserve the entire anterior ramus of the squamosal (E Tschopp, pers. obs., 2011) and it seems thus premature to include the distance as a phylogenetic character.

C57: Squamosal, posteroventral margin: smooth, or with short and blunt ventral projection (0); with prominent, ventrally directed ‘prong’ (1) (W11-37; modified; Fig. 13).

Comments. The original character description of Whitlock (2011a) was modified, and an additional binary character was added (see below) in order to describe better the state in Kaatedocus, where a short ventral projection of the squamosal is present.

C58: Squamosal, posteroventral margin: smooth, without ventral projection (0); ventral projection present (1) (W11-37; modified; Fig. 13).

Comments. A short projection is present in almost all preserved flagellicaudatan skulls. In contrast, most non-flagellicaudatan sauropods have smooth posteroventral margins of the squamosal.

C59: Parietal, contribution to posttemporal fenestra: present (0); absent (1) (W02-22; Fig. 14).

Comments. The absence of parietal contribution to the posttemporal fenestra is sometimes difficult to observe due to imperfectly preserved or distorted skulls. All diplodocid skulls have exoccipitals that bear a dorsolateral spur, which forms the dorsomedial end of the posttemporal fenestra (the ‘posttemporal process’ of Harris, 2006a). Additionally, most specimens have dorsally extended distal ends of the paroccipital processes, which curve back towards the exoccipital spur. These two prominences are interconnected by the squamosal in complete diplodocid skulls (CM 11161, E Tschopp, pers. obs., 2011).

C60: Parietal, portion contributing to skull roof, anteroposterior length/transverse width: wide, >50% (0); narrow, 7–50% (1); practically nonexistent, <7% (2) (New; Table S6). Ordered.

Comments. In some taxa, the posterior-most point of the fronto-parietal suture is located posterior to the supratemporal fenestra. The minimum values are compared in this ratio. Minimum anteroposterior length is measured between two parallel, transversely oriented lines intersecting the posterior-most point of the fronto-parietal suture and the anterior-most point of the concavity of the edge separating the dorsal portion of the parietal from the nuchal fossa.

C61: Parietal, distance separating supratemporal fenestrae: less than 1.5 times the width of the long axis of the supratemporal fenestra (0); at least 1.5 times the length of the long axis of the supratemporal fenestra (1) (W02-24; modified by M12-37; Table S7).

Comments. The original character states of Wilson (2002) left a gap (subequal, or double). The distance between the supratemporal fenestrae in many diplodocid specimens does not reach two times the maximum diameter of the fenestra, which led Mannion et al. (2012) to adjust the state boundaries. Specimens were remeasured where possible (Table S7), for others scorings of Wilson (2002) or Mannion et al. (2012) were used. The new measurements show that the ratios are often overestimated and that there seem to be three clusters of taxa (less than one: e.g., Giraffatitan; between one and 1.6 times: e.g., Kaatedocus; more than 1.6 times: e.g., Suuwassea). However, a more inclusive study of this character should be performed in order to recognize the most useful state boundaries for phylogenetic analyses. At the moment it seems wisest to retain the proposed version of Mannion et al. (2012).

C62: Parietal, posterolateral process, dorsal edge in posterior view: straight, and ventrolaterally oriented, so that the supratemporal fenestra is slightly facing posteriorly as well (0); convex, so that the postorbital and thus the supratemporal fenestra are not visible (1) (T13-43; Fig. 14).

Comments. The posterior view of the skull corresponds to the view parallel to the long axis of the occipital condylar neck, which was found to be oriented parallel to the lateral semicircular canal, thus indicating the neutral head position (Schmitt, 2012).

C63: Parietal, occipital process, dorsoventral height: low, subequal to less than the diameter of the foramen magnum (0); high, nearly twice the diameter of the foramen magnum (1) (W02-21; modified; Table S8).

Comments. Measurements are taken in strict posterior view (see above). Height is measured vertically between the dorsal-most and ventral-most extension of the occipital process, and the foramen magnum. In case of the occipital process, the dorsal- and ventral-most points are usually transversely shifted against each other. The measurements are therefore taken between horizontal lines intersecting the extremes. The state boundaries are tentatively set at 1.5, but more inclusive analyses would have to be undertaken in order to score this character adequately.

C64: Parietal, occipital process, distal end: ventrolaterally oriented, such that dorsolateral edge is straight or convex (0); curving laterally, such that dorsolateral edge becomes concave distally (1) (New; Fig. 14).

Comments. The distal end of the posterolateral process of the parietal of non-diplodocine flagellicaudatans curves outwards to meet the squamosal. This is not the case in the diplodocine skulls examined for this analysis.

C65: Parietal, distinct horizontal ridge separating dorsal from posterior portion: absent, transition more or less confluent (0); present, creating a distinct nuchal fossa below the ridge (1) (T13-44; wording modified; Fig. 15).

Comments. This character is best observed in oblique posterolateral view, if one does not have the specimens at hand. In the derived state, the transverse ridge caps the nuchal fossa dorsally, creating a distinct concavity below it. Given that small skulls appear to have this feature most expressed (AMNH 7530, CM 3452, SMA 0004), there is some possibility that the nuchal fossae become shallower during ontogeny.

C66: Postparietal foramen: absent (0); present (1) (U95; Fig. 5).

Comments. Postparietal foramina have been interpreted to be a dicraeosaurid synapomorphy (Whitlock, 2011a), but were recently shown to be present as well in Diplodocidae (Tschopp & Mateus, 2013b). The opening is located at the posteromedial corner of the two parietals, where they meet the supraoccipital. It might be associated with a vertical groove internally on the supraoccipital (Remes, 2006; see below), but additional CT studies would have to be performed in order to check for the presence or absence of this groove in specimens without the postparietal foramen. Many diplodocid specimens are damaged in this region of the skull, which makes it difficult to verify the presence of the foramen and impedes an evaluation of its distribution among flagellicaudatans. The definitive presence in Kaatedocus, and the unknown state in the two apatosaur skulls CM 11162 and YPM 1860 (due to crushing; E Tschopp, pers. obs., 2011), indicates that it might be plesiomorphic for Flagellicaudata, subsequently lost in Tornieria and Diplodocus.

C67: Paroccipital process (popr), posterior face: smooth/flat (0); with longitudinal ridge along popr body extending from dorsomedial to ventrolateral corners (1) (T13-46; Fig. 16).

Comments. Most of the specimens examined have a slightly convex posterior face of the paroccipital processes. However, few have such a distinct ridge as is present in Kaatedocus. In the latter, this ridge is accompanied by a rugose area at its dorsomedial origin. None of these structures are present in CM 11161, for example.

C68: Paroccipital process distal terminus: expanded vertically (0); not expanded (dorsal and ventral edges are subparallel) (1) (U98-38; modified; Fig. 14).

Comments. Upchurch (1998) included two morphologies in one character: the dorsoventral expansion, and the rounded or straight distal edge. The shape of the distal edge is difficult to assess qualitatively, because many specimens have slightly convex, or somewhat triangular lateral ends of the paroccipital process (e.g., Suuwassea ANS 21122, or Kaatedocus SMA 0004, Fig. 14). Therefore, the character description was limited to the distal expansion.

C69: Paroccipital process, distal end in lateral view: straight (0); curved (1) (New; Fig. 17).

Comments. Due to the slight posterior orientation of the paroccipital processes in many sauropod taxa, a strict lateral view of the skull does often not allow for an accurate coding of this character. Also, on pictures of articulated skulls it is often difficult to see the distal end of the paroccipital process well enough, because it is partly obscured by the squamosal. In most cases, a posterolateral instead of lateral view would thus be more helpful. Specimens, where the paroccipital processes were bent posteriorly during diagenesis should not be scored for this character because the pressure resulting in such a distortion likely also affected the curvature.

C70: Supratemporal fenestra: present, relatively large (anteroposterior diameter is at least 5% of occiput width) (0); absent, or greatly reduced (so that anteroposterior diameter is less than 5% of occipital width) (1) (W02-25; modified by M12-40).

Comments. Wilson (2002) proposed this feature as present/absent character, but Mannion et al. (2012) showed that one of Wilson’s (2002) derived taxa (Limaysaurus) actually has a supratemporal fenestra, although an extremely reduced one. Because this is a derived state of Rebbachisauridae, and because all diplodocid skulls show large openings, no additional measuring was done for this analysis.

C71: Supratemporal fenestra, maximum diameter: more than 1.2 times greatest diameter of foramen magnum (0); less than 1.2 times the greatest length of foramen magnum (1) (Y93-32; modified by M12-41).

Comments. Mannion et al. (2012) introduced the quantitative state boundaries to the original description (Yu, 1993). Basically, this character is an extension of the previous one, with the exception that Nigersaurus is impossible to score due to the complete absence of the supratemporal fenestra in this taxon. In addition to Limaysaurus, the quantitative boundaries of Mannion et al. (2012) also include the dicraeosaurids Dicraeosaurus and Amargasaurus, which have reduced supratemporal fenestra as well, but not to the extent shown by Rebbachisauridae. As stated above, the difference in relative size of the supratemporal fenestrae between the mentioned taxa and Diplodocidae is large, and thus no additional measurements were taken in order to test the boundaries proposed by Mannion et al. (2012).

C72: Supraoccipital, anterodorsal margin: internally concave, associated with a channel extending ventrally on the internal face (0); straight (1) (Remes, 2006; Fig. 18).

Comments. The channel was proposed by Remes (2006) as a distinguishing character between Tornieria and Dicraeosauridae, where the presence of the canal is coupled with the presence of a postparietal fenestra. However, as shown in Kaatedocus, these two features are not necessarily correlated. A separate coding for the two characters is thus justifiable. This is the first analysis to include this character.

C73: Supraoccipital, dorsal extension: high and vaulted, such that the dorsolateral edges are strongly sinuous (0); low, with the dorsolateral edges straight (1) (Remes, 2006; Fig. 14).

Comments. Remes (2006) used this character in order to distinguish Tornieria from Apatosaurus, but did not include it in his phylogenetic analysis. The present analysis is thus the first one to do so.

C74: Supraoccipital: sagittal nuchal crest: broad, weakly developed (0); narrow, sharp, and distinct (1) (W11-45; Fig. 19).

Comments. The nuchal crest lies on the midline of the supraoccipital, extending dorsoventrally. A narrow, sharp crest was previously thought to be a synapomorphy for Dicraeosauridae, but Tschopp & Mateus (2013b) showed that it also occurs in certain diplodocids.

C75: Supraoccipital, foramen close to contact with parietal: absent (0); present (1) (T13-52; Fig. 19).

Comments. This foramen is called an external occipital foramen by Balanoff, Bever & Ikejiri (2010) and is sometimes located entirely on the supraoccipital (Dicraeosaurus hansemanni MB.R.2379, Janensch, 1935), and in other cases on the suture with the parietal (Kaatedocus siberi SMA 0004, E Tschopp, pers. obs., 2010). Only taxa with well visible foramina are coded as apomorpic.

C76: Crista prootica, size: rudimentary (0); expanded laterally into dorsolateral process (1) (Salgado & Calvo, 1992; U95; Fig. 20).

Comments. Although diplodocids have a laterally protruding crista prootica (e.g., SMA 0011), only dicraeosaurids develop distinct lateral processes at the anteroventral ends of the crista prootica.

C77: Occipital condyle, articular surface: well offset from condylar neck (0); continuously grading into condylar neck (1) (New; Fig. 21).

Comments. Whereas in more basal sauropods the articular surface of the occipital condyle is usually well delimited, and offset from the condylar neck by a distinct ridge, diplodocids generally do not have such a clear distinction. The character states are most easily distinguished in dorsal view.

C78: Basioccipital, contribution to dorsal side of occipital condylar neck: present and broad, around 1/3 of entire dorsal side (0); reduced to absent (1) (Harris & Dodson, 2004; Fig. 14).

Comments. Harris & Dodson (2004) proposed the narrow contribution of the basioccipital to the dorsal face of the occipital condyle as characteristic for Suuwassea. A wider survey of the distribution of this character showed that the contribution of the basioccipital to the dorsal side of the occipital condylar neck is reduced in some diplodocid specimens as well.

C79: Basioccipital, distance from base of occipital condyle to base of basal tubera (best visible in lateral view): short, such that area is gently U-shaped in lateral view (0); elongate, with a flat portion between occipital condyle and basal tubera (1) (T13-54; wording modified; Fig. 17).

Comments. The distance is taken relative to the height of the basal tuber, creating a narrow U-shape or a shallow, wide concavity in lateral view (Fig. 17).

C80: Basioccipital depression between foramen magnum and basal tubera: absent (0); present (1) (W02-50; Fig. 22).

Comments. The depression is a concave area on the posterolateral sides of the basioccipital, which is different from the concavity on the posterior face of the basal tubera described in character 85.

C81: Basioccipital, pit between occipital condyle and basal tubera: absent (0); present (1) (M13-98; wording modified; Fig. 23).

Comments. Various pits can mark the area around the basal tubera: YPM 1860 bears one in the notch between the tubera (see below), and a second one on the basioccipital posterior to the tubera (which is the one described here). The basipterygoid recess is also located close by, but anterior to the basal tubera on the basisphenoid, instead of the basioccipital. Mannion et al. (2013) described this pit as a fossa on the posterior surface of the basal tubera, but this wording could be understood in a similar way as the concavity coded for in C85 herein. We therefore reworded the character to better delimit the character to the presence of this apparently blind foramen as seen in Fig. 23.

C82: Basal tubera: globular (0); box-like (1) (Whitlock, Wilson & Lamanna, 2010; Fig. 24).

Comments. Whitlock, Wilson & Lamanna (2010) used this character as one of the features distinguishing the juvenile diplodocid skull CM 11255 from Apatosaurus. It is herein used for the first time as a phylogenetic character.

C83: Basal tubera, breadth: <1.3 times (0); 1.3-1.85 times (1); >1.85 times occipital condyle width (2) (W02-49; modified; Table S9).

Comments. The character was initially defined without clear state borders, and only with two states (Wilson, 2002). Mannion (2011) suggested further subdivision of the character, based on a wider survey of this ratio among sauropods. Mannion’s (2011) table was here extended and the character state boundaries were modified following higher-level taxonomy and gaps in the distribution of the values.

C84: Basal tubera: distinct from basipterygoid (0); reduced to slight swelling on ventral surface of basipterygoid (1) (W11-53; Fig. 25).

Comments. The use of this character and its coding overlaps with an additional character proposed by Wilson (2002): ‘Basal tubera, anteroposterior depth: approximately 33%, or more, of dorsoventral height (0); sheetlike, less than 33% (normally around 20%) dorsoventral height (1).’ Whitlock’s (2011a) character is herein preferred because the directional terms used in Wilson (2002) are sometimes confusing due to varying orientations of the basal tubera of Diplodocoidea and non-diplodocoid sauropods.

C85: Basal tubera, shape of posterior face: convex (0); flat (1); slightly concave (2) (W11-54; modified by T13-63; Fig. 25).

Comments. The ‘posterior face’ of the basal tubera is herein intended to be the side facing the occipital condyle. The concavity described herein is different from the concavity sometimes present on the lateral side of the basioccipital (see above).

C86: Basal tubera, posteroventral face: continuous (0); marked by a distinct transverse ridge (1) (New; Fig. 24).

Comments. The surface of the basal tubera is usually regularly rugose, and without distinct structuring. SMA 0004, however, bears a distinct transverse ridge on the posteroventral face of its basal tubera.

C87: Basal tubera, longest axes: parallel (0); in an angle to each other, pointing towards the occipital condyle (1) (New; Fig. 26).

Comments. The character is to be coded based on a view perpendicular to the orientation of the basipterygoid processes. It is inspired by the character of Tschopp & Mateus (2013b) describing the anterior margin of the tubera as V- or U-shaped, which included two differing morphologies in the same character (orientation of the tubera and shape of the anterior margin). The two morphologies are here treated as different characters (see below). In some cases (e.g., CM 11162), the outline of the tubera is subtriangular, with a more or less right angle pointing posterolaterally. These cases were treated as apomorphic, because the longest distance follows the obliquely oriented hypotenuse of the triangle.

C88: Basal tubera, anterior edge: straight or convex (0); concave (1) (T13-64; Fig. 26).

Comments. The second of the two characters inspired by Tschopp & Mateus’ (2013b) character about the anterior margin of the basal tubera. The anterior edge is the one facing towards the basipterygoid processes, which in non-diplodocoid sauropods is oriented rather anteroventrally. In specimens with angled basal tubera (see above), the anterior margin is oriented obliquely.

C89: Basal tubera in posterior view: facing ventrolaterally (0); facing straight ventrally, forming a horizontal line (1) (T13-65; wording modified; Fig. 24).

Comments. Some specimens (in particular non-flagellicaudatans) have rounded basal tubera, which extend onto the lateral surface of the basioccipital. These are treated as plesiomorphic, because the line projecting through the medial- and lateral-most points of the tubera is oblique in these cases.

C90: Basal tubera, foramen in notch that separates the two tubera: absent (0); present (1) (T13-66; Fig. 23).

Comments. This foramen is one of three openings that can occur in this area (see above and below). However, the pit described in this character cannot be homologous to the other ones because it occurs together with the basipterygoid recess in HMNS 175 (Holland, 1906) and together with the basioccipital pit in YPM 1860 (E Tschopp, pers. obs., 2011).

C91: Basisphenoid/basipterygoid recess: absent (0); present (1) (W02-51; polarity reversed; Fig. 23)

Comments. The basipterygoid recess is a pit located anterior to the basal tubera, on the basisphenoid. Its absence was considered autapomorphic for Apatosaurus, representing a reversal to the plesiomorphic state in Sauropoda (Wilson, 2002). However, in his phylogenetic analysis, Wilson (2002) scored Apatosaurus as having a recess, sharing this state with basal sauropods like Shunosaurus. The character was organized as a presence/absence character, with the presence being plesiomorphic (Wilson, 2002). Assuming that the discussion of the autapomorphies is right, polarity of the character states was inverted herein. The basipterygoid recess might be confused with the pits located in the notch between the tubera or the one posterior to them (see above), so it is important to state that it lies anterior to the tubera, between the bases of the basipterygoid processes.

C92: Basipterygoid processes: widely diverging (>60°) (0); intermediate, 31°−60°(1); narrowly diverging (<31°) (2) (Y93-29; modified; Fig. 20; Table S10).

Comments. There are several modes to measure the angle between the processes, and no previous analysis defines how this angle should be measured. Here, divergence is measured between lines drawn from the basisphenoid center, where the bases of the basipterygoid processes meet, to the anteromedial-most point of the processes. This is preferably done in posterior or posteroventral view, perpendicular to the longitudinal axis of the processes. The present measuring technique yields slightly different results compared to earlier studies, but general trends are similar.

C93: Basipterygoid processes, orientation: directed more than 75° to skull roof (normally perpendicular) (0); angled less than 75° to skull roof (normally approximately 45°) (1) (McIntosh, 1990b; U98-41; modified; Table S11).

Comments. New numeric state boundaries were established, because a survey of diplodocoid braincases showed that there is more variety than previously recognized (Table S11). However, the difference was already recognized as taxonomically important by McIntosh (1990b). The angle is measured between the skull roof and a line through the center of the proximal and distal ends. This is important, especially because macronarian basipterygoid processes tend to curve backwards at their distal ends, thereby increasing the angle as measured here.

It is possible that this character is correlated with the large angle between the anterior and dorsal quadratojugal processes and the backwards inclination of the ventral ramus of the quadrate. This entire region is interconnected by the pterygoid, and the anterior shifting of the basisphenoid-pterygoid articulation due to the changed orientation of the basipterygoid processes might have been caused by, or the reason for the more anteriorly orientated ventral ramus of the quadrate, and therefore also the widening of the angle between the quadratojugal processes. However, because there is no evidence of correlation and no skulls are known of basal diplodocoid taxa that might show intermediate states, the separate characters are retained.

Furthermore, there is some indication that the character could be ontogenetically controlled: the two relatively small diplodocine skulls CM 3452 and SMA 0004 both have somewhat larger angles compared to larger specimens (Table S11), and lower angles in the quadratojugal. However, further studies are needed to decide if this is really ontogenetic, or if it could be taxonomically significant.

C94: Basipterygoid processes, ratio of length:basal transverse diameter: <4 (0); = or >4.0 (1) (W02-46; modified; Fig. 20; Table S12).

Comments. The character was initially defined as ratio of length to maximum basal diameter (Wilson, 2002). However, maximum basal diameter is often oriented dorsoventrally (at least in diplodocids), which means that one cannot take the measurements in a picture of the processes in ventral view only. Also, dorsoventral height changes considerably, and continuously, towards the base of the processes in some specimens (e.g., Dicraeosaurus hansemanni MB.R.2379; Janensch, 1935; Fig. 20). In lateral view, it is sometimes difficult to decide where exactly the base of the process is situated. Therefore, and because ventral views are obtainable more frequently than lateral views, the ratio length/basal transverse diameter is preferred herein. The dimensions should be measured perpendicular to each other. Wilson (2002) initially left a gap in the definition of the states (2 or less, 4 or more), which was corrected for by Mannion et al. (2012). However, as a more rigorous assessment of these ratios shows (Table S12), the state boundary should rather be set to four, the derived, elongate state resulting as a shared synapomorphy for Diplodocinae and Dicraeosauridae.

Measuring the basipterygoid processes in such a way leads to much higher elongation ratios for the holotype of Kaatedocus siberi (SMA 0004) than reported in its initial description (Tschopp & Mateus, 2013b). The low ratio also served as local autapomorphy for the genus (Tschopp & Mateus, 2013b). Following the results presented herein, this is most probably an artifact based on differing measurement protocols, because Tschopp & Mateus (2013b) compared length with dorsoventral height, which is the maximum basal diameter in SMA 0004 (Tschopp & Mateus, 2013b). The current measurements show that Kaatedocus is actually well in the range of Diplodocinae, which can easily be distinguished from Apatosaurus louisae CM 11162 (Table S12).

C95: Basipterygoid, area between the basipterygoid processes and parasphenoid rostrum: is a mildly concave subtriangular region (0); forms a deep slot-like cavity that passes posteriorly between the bases of the basipterygoid processes (1) (U95; U98-44; Fig. 20).

C96: Basipterygoid processes, orientation of proximal-most portions: same as central portion of shaft (0); parallel to each other, outwards curve of shaft happens only more anteriorly (1) (New; Fig. 27).

Comments. The development of this character is best seen in ventral view. In the derived state, the parallel portion of the basipterygoid processes are often interconnected dorsomedially by a thin sheet of bone. On the other hand, a similar sheet can also be present if the processes are entirely straight.

C97: Basipterygoid processes, distal end in anterior view: straight (0); curving laterally (1) (New; Fig. 22).

Comments. This character compares the distal end of the basipterygoid process with the central portion. It is thus different from the feature described in character 96.

C98: Basipterygoid processes, distal lateral expansion: absent (0); present (1) (New; Fig. 22).

Comments. Only abrupt distal expansions are coded as apomorphic. Gradually extending processes are treated as plesiomorphic.

C99: Parasphenoid rostrum, groove on dorsal edge: absent (0); present (1) (U95; U98-45; modified; Fig. 20).

Comments. Upchurch (1995) and Upchurch (1998) proposed the character combining the presence of a dorsal groove with the lateral shape of the rostrum, thereby implying that the dorsoventrally thin parasphenoid of diplodocoids would not bear dorsal grooves. However, a more detailed study of diplodocoids shows that the groove is actually present in most of them.

C100: Optic foramen: paired (0); unpaired (1) (Berman & McIntosh, 1978; S06-129; Fig. 28).

Comments. The optic foramen lies close to the midline, within the orbitosphenoid in most sauropod taxa. Generally, the right and left foramina are separated medially by a narrow bony bridge, which is absent in some diplodocoid specimens (e.g., Suuwassea, Harris, 2006a). Sander et al. (2006) were the first to include the character in a phylogenetic analysis.

C101: Palatobasal contact, shape: pterygoid with small facet (0); dorsomedially orientated hook (1) (W02-36; modified by T13-67; Fig. 29).

Comments. Tschopp & Mateus (2013b) deleted a third state from the original character, which describes the specific rocker-like morphology of this region in nemegtosaurid sauropods (Wilson, 2002). Because no taxon of this clade is included, the additional state is redundant here.

C102: Pterygoid, transverse flange (i.e., ectopterygoid process) position: between orbit and antorbital fenestra (0); anterior to antorbital fenestra (1) (U95; Fig. 9).

Comments. The transverse flange of the pterygoid connects to the maxilla through the ectopterygoid (Upchurch, Barrett & Dodson, 2004).

C103: Vomer, anterior articulation: maxilla (0); premaxilla (1) (W02-42; polarity reversed; Fig. 9).

Comments. Polarity was reversed compared to Wilson’s (2002) character due to the limited taxon sampling.

C104: Dentary, anteroventral margin shape: gently rounded (0); sharply projecting triangular process or ‘chin’ (1) (U98-58, modified by W02-56; Fig. 30).

Comments. Usually considered a flagellicaudatan synapomorphy, some specimens of Camarasaurus also show a weak ventral expansion at the anterior extreme of the lower jaw. However, this never reaches the chin-like state present in Diplodocus, and Camarasaurus is thus included in the plesiomorphic state here.

C105: Dentary, cross-sectional shape of symphysis: oblong or rectangular (0); subtriangular, tapering sharply towards ventral extreme (1); subcircular (2) (W11-60; Fig. 30).

Comments. Diplodocids have ventrally tapering symphyses, but they do not taper to a point as in dicraeosaurids (Whitlock & Harris, 2010) and were thus scored as plesiomorphic.

C106: Dentary, tuberosity on labial surface near symphysis: absent (0); present (1) (Whitlock & Harris, 2010; reworded by W11-57; Fig. 31).

Comments. This character was originally proposed by Whitlock & Harris (2010) to unite Suuwassea and Dicraeosaurus.

C107: Dentary, anterolateral corner: not expanded laterally beyond mandibular ramus (0); expanded beyond lateral mandibular ramus (1) (W11-59; Fig. 31).

Comments. The derived state of this character describes the extreme case of character 112. To date, it is only known in the rebbachisaurid Nigersaurus (Sereno et al., 2007).

C108: Mandible, coronoid eminence: strongly expressed, clearly rising above plane of dentigerous portion (0); absent (1) (W11-62; Fig. 30).

Comments. Some diplodocids have dorsally expanded coronoid areas, but they do not reach above the plane of the dentigerous portion.

C109: Surangular foramen: absent (0); present (1) (New; Fig. 32).

Comments. The location of the surangular foramen can vary in different taxa. Usually, it is situated in the anterodorsal portion, but in some cases it is shifted posteriorly.

C110: External mandibular fenestra: present (0); absent (1) (McIntosh, 1990b; R93-3; Fig. 32).

Comments. The presence is a retained plesiomorphy, shared with early sauropodomorphs (Wilson, 2002).

C111: Snout shape in dorsal view: premaxilla-maxilla index (PMI; Whitlock, Wilson & Lamanna, 2010) <67% (0); 67-85% (1); >85% (2) (U98-1; W11-64; modified; Table S13). Ordered.

Comments. In order to avoid gaps, an intermediate state was added to Whitlock’s (2011a) version. The state boundaries were chosen following high-level phylogenetic differences. Measurements taken on photographs from slightly different angles of the skulls CM 3452, 11161, 11162, and SMA 0011 show that the orientation of the skull has a relatively high influence on the measured PMI (Table S13). In order to avoid this, the same measurements were taken in more than one picture of the same skulls, where possible. In future, one should check and remeasure this ratio in all diplodocid skulls, making sure that they are always taken in exactly the same orientation. Best results are to be expected with the ventral maxillary edge oriented horizontally.

Whitlock, Wilson & Lamanna (2010) reported that the snout becomes more squared during ontogeny in diplodocids. It might thus be possible that more juvenile specimens become artificially grouped closer to more basal taxa when including this character.

Teeth

C112: Shape of tooth row in occlusal view: follows curvature of dentary (0); anterolateral corner of tooth row displaced labially (1) (Whitlock & Harris, 2010; Fig. 31).

Comments. In dicraeosaurids, the tooth row seems to be the main responsible for the squared appearance of the lower jaw. The ventral portions of the dentary would be much more rounded (Whitlock & Harris, 2010). The diplodocid AMNH 969 has a similar development as Suuwassea.

C113: Tooth rows, length: restricted anterior to orbit (0); restricted anterior to antorbital fenestra (1); restricted anterior to subnarial foramen (2) (G86; modified by W11-65; Fig. 1). Ordered.

Comments. In order to score this character, the skull should be held with the ventral margin of the maxilla oriented horizontally. The tooth row is usually more anteriorly restricted in the lower jaw than in the maxilla. Here, the maxillary tooth row is used as a reference. As for the snout shape, the anterior restriction of the tooth row also was interpreted as juvenile feature (Whitlock, Wilson & Lamanna, 2010).

C114: Dentary teeth, number: greater than 17 (0); 10-17 (1); 9 or fewer (2) (W98-67; modified by C12b-96; Table S14). Unordered.

Comments. Carballido et al. (2012b) added a third state to distinguish Demandasaurus and Suuwassea from other sauropod specimens. Given that the derived state is ambiguous, it is more parsimonious to leave the character unordered.

C115: Replacement teeth per alveolus, number: three or fewer (0); four or more (1) (W02-74, modified by W11-71).

Comments. The number of replacement teeth varies between the tooth-bearing bones of the same individual (D Schwarz, pers. comm., 2012). However, maximum number of replacement teeth is still informative, and therefore the character was retained.

C116: Teeth, crown-to-crown occlusion: present (0); absent (1) (W98-35; polarity reversed by W11-66).

C117: Teeth, wear facets shape: v-shaped (0); planar (1) (W98-36; modified; Figs. 33 and 34).

Comments. The initial character (Wilson & Sereno, 1998) was first adapted by Sereno et al. (2007), in order to include the paired planar facets of Nigersaurus. Here, the shape and number of wear facets are considered independent characters (see character 118), because they code for varying morphology or processes of food intake.

C118: Teeth, occlusal pattern: paired wear facets (0); single facet (1) (W98-36; modified; Fig. 34).

Comments. See character 117.

C119: Teeth, SI values for tooth crowns: <3.4 (0); 3.4 or greater (1) (McIntosh, 1989; U98-69; modified; Table S15).

Comments. The SI value describes the slenderness of the teeth. It was defined as crown length/mesiodistal width (Upchurch, 1998). The state borders were changed, following large gaps apparently corresponding to higher-level taxonomy (Table S15).

C120: Tooth crowns, orientation: aligned slightly anterolingually, tooth crowns overlap (0); aligned along jaw axis, crowns do not overlap (1) (W98-34; polarity reversed by W11-68; Fig. 32).

C121: Tooth crowns, cross-sectional shape at midcrown: D-shaped (0); cylindrical (1) (R93-7; modified by W98-32; Fig. 35).

Comments. Unworn diplodocoid teeth often have ellipsoid cross-sections. However, this is different from the spatulate non-diplodocoid teeth as e.g., typical for Camarasaurus. Teeth of the latter genus have a slightly concave lingual face, unlike the convex surface of diplodocoids. In the absence of nemegtosaurid titanosaurs, which show similarly shaped teeth (Upchurch, 1999; Wilson, 2005), the derived state results as an unambiguous synapomorphy of Diplodocoidea.

C122: Teeth, orientation relative to long axis of jaw: perpendicular (0); oriented anteriorly (procumbent) (1) (G86, U98-72; Fig. 32).

Comments. Tooth orientation is best recognized in the posterior-most teeth in the maxilla and dentary.

C123: Teeth, longitudinal grooves on lingual aspect: absent (0); present (1) (W02-76; Fig. 33).

Comments. Wilson (2002) initially scored only rebbachisaurids with the derived state. However, several non-diplodocoid taxa with spatulate teeth actually have a midline ridge on the lingual face of their teeth, creating two grooves mesially and distally to it (e.g., Osborn & Mook, 1921; Ouyang & Ye, 2002). Consequently, these taxa are scored as derived here as well.

C124: Teeth, thickness of enamel asymmetric labiolingually: absent (0); present (1) (W11-74; Fig. 35).

Comments. This feature can be observed easily in wear facets or cross-sections.

C125: Teeth, marginal denticles: present (0); absent (1) (McIntosh, 1990b; U98-66; Fig. 33).

Comments. There is some morphological variation in the location of the denticles (Carballido et al., 2012b), but because no diplodocid shows denticles, this simplified version of the character is used herein.

Cervical vertebrae

C126: Presacral neural spines, bifurcation: absent (0); present (1) (McIntosh, 1989; W02-85, 89; modified; Table S16).

Comments. Wilson (2002) divided this character into the different regions, where the bifurcation can be present. As a result, taxa with unbifurcated neural spines are coded several times for the same state. In the present analysis, presence of bifurcation and the first bifid element are treated as two different characters (see character 140).

C127: Number of cervical vertebrae: <13 (0); 14–15 (1); 16 or more (2) (McIntosh, 1990b; W98-37; modified; Table S17). Unordered.

Comments. The character is used in various versions in different phylogenetic analyses (Upchurch, 1998; Wilson & Sereno, 1998; Whitlock, 2011a), depending on their specific focus. Herein, the states are adjusted to fit the included taxa, excluding redundancy. Only one diplodocid specimen preserves a complete neck (Apatosaurus louisae CM 3018), and even here, the possibility of missing elements cannot be ruled out entirely, due to gaps between certain cervical vertebrae as they were found (McIntosh, 2005). A second specimen (Diplodocus carnegii CM 84) lacks the atlas, and seems otherwise complete, although the same concerns exist as for CM 3018 (McIntosh, 2005). However, as the more anterior and posterior elements in these cases fit well together, we followed McIntosh (2005) in assuming that no vertebra was lost at the position of these gaps in CM 84 and 3018. McIntosh (2005) suggested that Barosaurus had 16 cervical vertebrae, instead of 15 as Apatosaurus and Diplodocus. The assumption was primarily based on the fact that AMNH 6341 only has nine dorsal vertebrae, and that the neosauropod presacral column generally consists of 25 elements (McIntosh, 2005). Because none of the Barosaurus specimens preserves an entire neck, none of the Barosaurus OTUs can be coded for this character. The inability to code incomplete specimens might be circumvented by using additive binary characters (Upchurch, 1998). However, this would imply that the corresponding multistate character is continuous (Wilson, 2002), which means that the number of cervical vertebrae could not increase directly by more than one element during speciation. Given that the contrary is shown to be possible in dorsal and sacral vertebrae of mice (Wellik & Capecchi, 2003), it seems reasonable to argue that the same accounts for sauropod cervical vertebrae. The character is thus treated as unordered herein. This also indicates that ‘analysis 1’ of Mannion et al. (2012), where these characters are treated as unordered, should be preferred over ‘analysis 2.’

C128: Cervical vertebrae width to height ratio: less than 0.5 (0); 0.5–1.5 (1); more than 1.5 (2) (U04b-1; modified; Table S18). Unordered.

Comments. Upchurch, Tomida & Barrett (2004, p. 105) defined the ratio as follows: “Height is measured from the top of the neural spine to the ventral surface of the centrum. Width is defined as the distance between the distal tips of the diapophyses.” A third state was added (less than 0.5) because derived dicraeosaurids have a distinctly lower ratio compared to other flagellicaudatans. Given that outgroups are scored for state 1, this character is left unordered.

C129: Cervical pneumatopores (pleurocoels): absent (0); present (1) (McIntosh, 1990b; U95; Fig. 36).

Comments. McIntosh (1990b) already used this character to distinguish advanced sauropods from the most basal forms, but Upchurch (1995) was the first to include it into a phylogenetic analysis.

C130: Cervical centra, internal pneumaticity: absent (0); present with single and wide cavities (1); present, with several small and complex internal cavities (2) (W98-102; modified by C12b-120; Fig. 37).

Comments. Introduced as a character by Wilson & Sereno (1998), only Wedel, Cifelli & Sanders (2000) and Wedel (2003) analyzed the distribution of this feature in detail. Carballido et al. (2012b) divided the original character, which did not discriminate between cervical and dorsal vertebrae (Wilson & Sereno, 1998).

C131: Cervical vertebrae, small fossa on posteroventral corner: absent (0); shallow, anteroposteriorly elongate fossa present, posteroventral to pleurocoel (1) (W11-83; Fig. 36).

Comments. Kaatedocus siberi SMA 0004, AMNH 7530, and the apatosaurines YPM 1980 and AMNH 460 have shallow depressions at the same place, but they do not create distinct fossae as in Barosaurus or Diplodocus (see Hatcher, 1901; McIntosh, 2005), and are thus coded as plesiomorphic.

C132: Cervical centra, midline keels on ventral surface: prominent and plate-like (0); reduced to low ridges (1) (U98-83; modified; Fig. 38).

Comments. Because the presence or absence is already coded in subsequent characters, the complete absence is here excluded from the original character description (Upchurch, 1998), and taxa without ventral ridges are scored as unknown.

C133: Cervical vertebrae, longitudinal sulcus on ventral surface: absent (0); present (1) (U95, U98-84; Fig. 38).

Comments. Due to the lateroventral projecting cervical parapophyses of Apatosaurus, cervical vertebrae of this genus have a concave anterior portion of the ventral surface. However, this is the case in almost all sauropod taxa, and therefore only specimens with transversely concave ventral surfaces throughout the entire length of the centrum are herein scored as apomorphic.

C134: Cervical vertebra, posterior projection on transverse processes: present (0); absent (1) (R09-78; polarity reversed; Fig. 39).

Comments. A distinct, triangular posterior projection marks the transverse process of Spinophorosaurus and many diplodocines. Posteriorly convex transverse processes are not considered projections. Due to reduced taxon sampling, the character polarity of the original version (Remes et al., 2009) was inverted here.

C135: Cervical vertebrae, posterior extension of posterior centrodiapophyseal lamina: is anteriorly restricted (0); reaches below posterior end of neural canal (1) (New; Figs. 36 and 40).

Comments. Apatosaurus specimens appear to have a consistently more developed pcdl compared to Diplodocinae. The only apatosaur specimen with an anteriorly restricted pcdl is the juvenile holotype of Elosaurus parvus, CM 566. However, because the development of vertebral laminae has previously been linked with ontogeny (Schwarz, Frey & Meyer, 2007b; Carballido & Sander, 2014), the anteriorly restricted pcdl in CM 566 might be an ontogenetic feature. Articulated cervical series (e.g., Apatosaurus louisae CM 3018, Diplodocus carnegii CM 84, Kaatedocus siberi SMA 0004) show that this character is stable throughout the column, and can thus be used in all cervical sections.

C136: Cervical vertebrae, short second posterior centrodiapophyseal lamina ventral to the one uniting with the dorsal shelf of the diapophysis: absent (0); present (1) (New; Fig. 41).

Comments. A short accessory pcdl appears to be linked with the bifurcation of the pcdl in more posterior elements in SMA 0011. However, a bifurcated pcdl also occurs in some apatosaur specimens, which do not have an additional pcdl in more anterior elements (e.g., UW 15556; Gilmore, 1936), and therefore, these morphologies are treated as independent characters.

C137: Cervical vertebrae, foramen on dorsal side of postzygodiapophyseal lamina, just anterior to base of neural spine process: absent (0); present (1) (Remes, 2007; Fig. 41).

Comments. Distinct foramina in the sdf are usually considered typical for brachiosaurids, and their presence in Australodocus was therefore one of the reasons why Whitlock (2011c) reinterpreted Australodocus bohetii as a titanosauriform, instead of a diplodocine as initially proposed (Remes, 2007). However, Barosaurus sometimes shows small foramina in similar positions (YPM 429, E Tschopp, pers. obs., 2011), but they are usually less prominent. The putative juvenile Brachiosaurus specimen SMA 0009 does not have such foramina, but because the development of pneumatic structures appears to be ontogenetically controlled (Schwarz et al., 2007; Carballido et al., 2012a), this might be explained as such.

C138: Cervical vertebrae, epipophysis: reduced or absent (0); pronounced, forming a distinct projection above the postzygapophysis (1) (R09-80; modified; Fig. 41).

C139: Cervical vertebrae, pneumatized epipophyses: absent (0); present (1) (New; Fig. 42).

Comments. The pneumatic foramen can be situated anteriorly as in Diplodocus carnegii (CM 84, 94, E Tschopp, pers. obs., 2011), or posteriorly as in Barosaurus lentus YPM 429 (E Tschopp, pers. obs., 2011).

C140: Cervical neural spines, bifurcation, if present, anterior extension within column includes: CV 3 (0); all mCV (1); posterior mCV (2); only pCV (3) (R93-9; modified; Table S16). Ordered.

Comments. Taxa with unbifurcated neural spines are scored as unknown. The subdivision into anterior, mid-, and posterior cervical vertebrae depends on the number of elements in the column (Table 3). Absolute numbers other than CV 3, which is the first postaxial cervical element, would thus be misleading and are avoided here.

C141: Cervical vertebrae, unbifurcated neural spines in anterior/posterior view: with parallel lateral edges or converging (0); distal end expanded laterally (1) (New; Fig. 43).

Comments. The real distribution of this character within Diplodocidae is difficult to assess to date, because there are only a few specimens reported that preserve complete neural spines of anterior, unbifurcated neural spines.

C142: Cervical vertebrae, summits of bifid neural spines: are laterally compressed (0); are rounded (1) (U04b-7; Fig. 39).

Comments. The derived state of this character is shared by some apatosaur specimens and Suuwassea. The spine summits in most other taxa with bifurcated spines are generally anteroposteriorly elongate and transversely compressed, resulting in narrow sheets of bone. In Suuwassea as well as in some apatosaur specimens, the lateral edge of the spine summit is distinctly convex, producing a semi-circular outline. Some other taxa (e.g., Kaatedocus; Tschopp & Mateus, 2013b) have medial ridges connecting the summit with the base, but these are always relatively shallow, and do not form rounded outlines. Taxa with unbifurcated neural spines are scored as unknown.

C143: Proatlas, distal end: broadly rounded (0); narrow and elongate, almost pointed (1) (New; Fig. 44).

C144: Atlantal intercentrum, anteroventral lip: absent, anterior edge of intercentrum straight in lateral view (0); present, anterior edge of intercentrum concave (1) (W02-79; modified; Fig. 45).

Comments. Initially regarded as flagellicaudatan synapomorphy (Wilson, 2002), an anteroventral lip is now known to occur in Mongolosaurus as well (Mannion, 2011). Following the original description of the character states (Wilson, 2002: intercentrum shape in lateral view: rectangular or ventrally longer than dorsally), Camarasaurus and other non-flagellicaudatan taxa also would have to be scored as apomorphic. However, they do not show a distinct anteroventral lip, resulting in a strongly concave anterior edge of the intercentrum, when seen in lateral view.

C145: Atlantal intercentrum, ventral surface, foramen between posterior ventrolateral processes: absent (0); present (1) (New; Fig. 45).

C146: Atlantal neurapophyses, anteromedial process: weakly developed (0); well-developed and distinct from posterior wing (1) (New; Fig. 46).

Comments. The anteromedial process corresponds to the prezygapophyses of more posterior elements. It articulates with the posterior end of the proatlas. In Kaatedocus, this process is relatively short transversely, and curves gradually into the posterior process, whereas in SMA 0011 and AMNH 969 the anteromedial process is distinct and at least as wide transversely as long anteroposteriorly.

C147: Atlantal neural arch, small subtriangular, laterally projecting spur at base: absent (0); present (1) (New; Fig. 46).

Comments. When present, this spur is located at the base of the neurapophysis, opposite the anteromedial process, and much smaller. It is also present in some, but not all, Camarasaurus specimens (Ikejiri, 2004).

C148: Atlantal neurapophyses, posterior wing: gradually tapering along its length (0); of subequal width along most of its length (1) (New; Fig. 46).

Comments. The posterior wing of the neurapophysis articulates with the prezygapophysis of the axis.

C149: Atlantal neurapophyses, posterior wing: without foramen (0); with foramen (1) (Wilson, 2002; W11-85; wording modified; Fig. 46).

Comments. Wilson (2002) proposed the presence of such a foramen as an autapomorphy of Apatosaurus, and it was included as character in the phylogenetic analysis of Whitlock (2011a). Due to the small number of preserved atlantal neurapophyses, only one specimen can currently be positively assigned to the apomorphic state (Apatosaurus louisae CM 3018). It could thus also represent a species autapomorphy, instead of being valid for the entire genus.

C150: Axial centrum, pneumatic fossae on ventrolateral edges, posterior and adjacent to parapophyses: absent (0); present (1) (New; Fig. 47).

Comments. Many specimens have a well-developed median keel on their ventral surfaces. In lateral view, this sometimes appears as a bifurcation of the ventrolateral edge, although this is not the case. The apomorphic state of the character proposed herein only includes fossae bordered by ridges that originate at the parapophysis anteriorly.

C151: Axis, prespinal lamina: of constant width (0); developing a transversely expanded, knob-like tuberosity at its anterior end (1) (New; Fig. 48).

C152: Axis, postspinal lamina: absent (0); present (1) (Harris & Dodson, 2004; Fig. 47).

C153: Axis neural spine: projects beyond posterior border of centrum (0); terminates in front of or at posterior border of centrum (1); is restricted anterior to postzygapophyseal facets (2) (New; Fig. 48). Ordered.

Comments. Due to intermediate morphologies, this character is treated as ordered.

C154: Anterior cervical vertebrae, total height/centrum length ratio: <0.9 (0); 0.9–1.2 (1); >1.2 (usually around 1.5) (1) (W11-87; modified; Table S19). Unordered.

Comments. Total height is herein measured between the ventral-most expansion of the centrum (usually the parapophysis or posterior cotyle) and the highest point of the neural spine. A third state was added in order to distinguish apatosaurs from Diplodocus. Given the high amount of changes in ratios during evolution, as indicated by the analysis, the character is left unordered.

C155: Cervical vertebrae 2 and 3, centrum length: moderate length increase, CV3 <1.3 × CV 2 (0); length increases considerably CV 3 at least 1.3 × CV 2 (1) (Russell & Zheng, 1993; Table S20).

Comments. Even though this does not seem to follow higher-level taxonomy, there are two groups with ratios well separated from each other (Table S20). The state boundaries are therefore set in order to distinguish between these two groups.

C156: Anterior cervical vertebrae, posterior edge of anterior condyle: anteriorly inclined (0); posteriorly inclined (1) (New; Fig. 49).

Comments. This character is strictly applicable to anterior cervical vertebrae. In SMA 0011, which has apomorphic anterior vertebrae, CV 6 and more posterior elements show the usual anteriorly inclined edge.

C157: Anterior cervical centra, pleurocoels: single (0); subdivided (1) (New; Fig. 49).

Comments. The subdivision of the pleurocentral cavity is sometimes regarded as ontogenetically controlled (Schwarz, Frey & Meyer, 2007b; Carballido & Sander, 2014). However, given that the completely mature anterior cervical vertebrae (sensu Carballido & Sander, 2014) of the Kaatedocus siberi holotype SMA 0004 have undivided pleurocoels, in contrast to the still immature vertebrae of other specimens like SMA 0011 (see above), at least some taxonomic differences are likely.

C158: Anterior cervical vertebrae, pleurocoel extending onto dorsal surface of parapophysis: absent (0); present (1) (U98-86; modified by W11-88; polarity reversed; Fig. 49).

Comments. Upchurch (1998) distinguished between continuous extensions or fossae that are separated from the main anterior pneumatic fossa or pleurocoel by a transverse ridge. The latter distinction was abandoned by Whitlock (2011a), who instead divided the character into the different regions (anterior and mid- and posterior cervical vertebrae, see below). Character polarity was herein reversed because basal outgroups used in the present analysis do have expanded pleurocoels.

C159: Anterior cervical vertebrae, longitudinal ridge on ventral surface: present (0); absent (1) (U98-83; modified).

Comments. The ventral ridge (if present) can have various morphologies in diplodocid specimens, which is accounted for in other characters of this analysis. In addition to the original version of Upchurch (1998; character 132 herein), a strict presence–absence character was included for both anterior and mid- and posterior cervical vertebrae in the present analysis. The subdivision is necessary because in some specimens, a ventral keel only occurs in anterior elements (ANS 21122, SMA 0011, Tate-001). This indicates that incomplete necks without ventral keels on posterior cervical vertebrae might still bear midline ridges anteriorly. For the various developments of the keels see Fig. 38, which shows mid- and posterior cervical vertebrae, but the morphology is the same in anterior elements.

C160: Anterior cervical vertebrae, paired pneumatic fossae on ventral surface: absent (0); present (1) (W11-89).

Comments. Like the ventral keel, the paired pneumatic foramina are sometimes restricted to the anterior cervical vertebrae (e.g., in SMA 0011, see above). Whereas the presence of paired pneumatic foramina imply the presence of a ventral keel, this does not apply the other way around, as shown by the anterior cervical vertebrae of Kaatedocus SMA 0004 (Tschopp & Mateus, 2013b). The characters are therefore retained as independent. The morphology of the foramina is equal in anterior and mid- and posterior cervical vertebrae, where present (see Fig. 38). In our analysis, paired pneumatic foramina only occur at the anterior end of the ventral surfaces. However, given that paired fossae in the posterior cervical vertebra of Dinheirosaurus lourinhanensis ML 414 occur at the posterior end of the ventral surface, we refrained from restricting the character definition to anteriorly placed foramina.

C161: Anterior cervical vertebrae, prespinal lamina: absent (0); present (1) (C12b-121; Figs. 43 and 49).

Comments. In some diplodocid specimens, it appears that the prespinal lamina in undivided vertebrae gives rise to the median tubercle in divided, more posterior elements. However, given the presence of a prespinal lamina in Camarasaurus (Madsen, McIntosh & Berman, 1995), which does not have a median tubercle between bifurcated neural spines, these two characters should be treated as independent.

C162: Anterior and mid-cervical centra, pleurocoel pierced by one or two large, rounded foramina around centrum midlength: absent (0); present (1) (New; Fig. 50).

Comments. Such a foramen is absent in the anterior-most elements, but very distinct in CV 5 or 6 of SMA 0011, whereas it disappears again by CV 8 or 9. In SMA 0011, these foramina are situated at the anterior end of the posterior pneumatic fossa. Taxa where CV 5 to 7 or 8 are not preserved, and other elements do not show such a development, are scored as unknown. Similarly distinct, rounded foramina are only present in Supersaurus (Lovelace, Hartman & Wahl, 2007), and Australodocus (Remes, 2007; Whitlock, 2011c).

C163: Anterior and mid-cervical vertebrae, spinoprezygapophyseal lamina, development at base of prezygapophyseal process: distinct (0); reduced to broad ridge or totally interrupted (1). (T13-103; wording modified; Fig. 51).

Comments. The character was clarified in order to specify that the reduction to a ridge and the interruption of the sprl are restricted to the base of the prezygapophyseal process. Otherwise one could understand that the reduction to a ridge would affect the entire sprl, which is not what was intended to code for with this character initially.

C164: Anterior and mid-cervical neural spines height: high (project well above the level of postzygapophyses) (0); low (terminates level with postzygapophyses) (1) (U04b-8; modified; Fig. 41).

Comments. This character is similar to character 168. It was added because it includes anterior cervical vertebrae, which are different in height among diplodocids and within Diplodocinae, and because it would have differing state boundaries, if it would be treated numerically.

C165: Anterior and mid-cervical neural spines, dorsoventrally elongate coel on lateral surface: absent (0); present (1) (M12-99; modified; Fig. 50).

Comments. The presence of a dorsoventrally elongate fossa in the spinodiapophyseal fossa is usually used as derived character for posterior cervical vertebrae only (Mannion et al., 2012). However, there are differences in anterior and mid-cervical neural arches as well, which appear to be phylogenetically significant.

C166: Mid-cervical centra, anteroposterior length/height of posterior face: 2.5–3.2 (0); 3.3–4.4 (1); 4.5+(2) (U95; modified; Table S21).

Comments. Elongation index as used herein is measured following the protocol of Wilson & Sereno (1998: total centrum length/height posterior cotyle). The mean elongation index is used for this metric. Tornieria specimen k is scored ‘2’ because the centrum length to width ratio is very high (5.4; Remes, 2006), and thus a high EI as used herein can be expected with confidence.

C167: Mid-cervical pre-epipophyses anterior extreme: about the same as prezygapophyseal facet (0); projects considerably anterior to articular facet, forming a distinct spur (1) (Sereno et al., 1999; Fig. 51).

Comments. A distinct anterior extension of the pre-epipophysis was used as an autapomorphy for Australodocus bohetii within Diplodocidae (Remes, 2007). However, it has been shown to be present in Kaatedocus as well as in some non-diplodocid sauropods (Sereno et al., 1999; Ksepka & Norell, 2006; Tschopp & Mateus, 2013b). Taxa without pre-epipophyses are scored as unknown.

C168: Mid-cervical neural spine height: considerably shorter than height of neural arch, <0.45 (0); subequal to height of neural arch, 0.45–1.6 (1); considerably higher than neural arch, >1.6 (2) (R05-69; modified; Table S22). Unordered.

Comments. Neural arch height is measured in a vertical line from the centrum to an imaginary line connecting the dorsal edges of the postzygapophyses, and neural spine height from dorsal edge of the postzygapophyses to the spine summit. The centrum is oriented such that the ventral floor of the neural canal is horizontal. The majority of the ratios were measured from photographs or figures in lateral view. As exemplified by CV 6 of Suuwassea ANS 21122, this approach can yield major differences depending on slight changes in perspective (or left and right lateral views; CV 6 of ANS 21122 has ratios ranging from 0.91–1.27; Table S22). Although such differences are partly avoided by using mean ratios, it would be unwise to use closely spaced numerical state boundaries in this case. Therefore, only two steps were regarded as sufficiently objective and phylogenetically significant. The character was left unordered due to diverging evolutionary trends.

C169: Mid-cervical neural spines, orientation: vertical (0); anteriorly inclined (1) (R05-68; Fig. 52).

Comments. The neural spine is interpreted to be anteriorly inclined, when the anterior end of the summit reaches further anterior than the posterior-most point of the sprl.

C170: Mid-cervical vertebrae, angle between postzygodiapophyseal and spinopostzygapophyseal laminae: acute (0); right angle (1) (R05-67; Fig. 50).

Comments. Angles are measured between lines connecting the posterior-most point of podl and spol (often the epipophyses) with their opposing ends.

C171: Mid- and posterior cervical centra, pleurocoels: single without division (0) divided by a bone septum, resulting in an anterior and a posterior lateral excavation (1); divided in three or more lateral excavations, resulting in a complex morphology (2) (C12b-115; modified; Fig. 36).

Comments. The original character (Carballido et al., 2012b) includes a fourth character state, which describes the shallow posterior pneumatic fossa. As such, it overlaps with character 172, introduced by Whitlock (2011a). Furthermore, subdivision of the pleurocoel is not correlated with the depth of the single pneumatic fossae in diplodocids. Therefore, the fourth state was omitted here.

C172: Mid- and posterior cervical vertebrae, pneumatization of lateral surface of centra: large, divided pleurocoel over approximately half of centrum (0); reduced, large fossa but sharp-bordered coel, if present, restricted to area above parapophysis (1) (W11-81; Fig. 40).

Comments. Taxa with single pleurocoels are scored as unknown.

C173: Mid- and posterior cervical vertebrae, pleurocoel extending onto dorsal surface of parapophysis: present (0); absent (1) (U98-86; modified by W11-95; Fig. 36).

C174: Mid- and posterior cervical vertebrae, longitudinal ridge on ventral surface: present (0); absent (1) (New).

C175: Mid- and posterior cervical vertebrae, ventral keel: single (0); bifid, connects posterolaterally to the ventrolateral edges of the centrum (1) (New; Fig. 38).

Comments. Taxa without ventral keels are scored as unknown.

C176: Mid- and posterior cervical vertebrae, paired pneumatic fossae on ventral surface, separated by ventral midline keel: absent (0); present (1) (New; Figs. 38 and 53).

Comments. Usually, these fossae are situated anteriorly between the parapophyses, separated by a ventral keel. Some apatosaur specimens (e.g., YPM 1861, E Tschopp, pers. obs., 2011) show paired pneumatic fossae located posterior to the parapophyses, facing ventrolaterally, and not separated by a keel. This morphology is considered different, and accounted for in character 177.

C177: Mid- and posterior cervical vertebrae, lateral edge posterior to parapophysis: continuous (0); marked by a deep groove extending anteroposteriorly along the edge (1) (New; Fig. 53).

Comments. This groove results in the presence of two distinct laminae or ridges extending from the parapophysis posteriorly.

C178: Mid- and posterior cervical centra, rugose tuberosity on anterodorsal corner of lateral side: absent (0); present (1) (T13-120; modified; Fig. 52).

Comments. The character description was extended to mid-cervical vertebrae in order to include Suuwassea emilieae. In the latter, the distinct rugose tubercles appear in mid-cervical vertebrae, whereas in Kaatedocus siberi, mid-cervical vertebrae only have very shallow tubercles. An additional character for serial variation is avoided because it could only be scored for these two taxa and would thus not be phylogenetically significant.

C179: Mid- and posterior cervical centra with longitudinal flanges in the lateroventral edge on the posterior part of the centrum: absent (0); present (1) (T13-113; Fig. 38).

Comments. These flanges are mainly responsible for the posterior portion of the ventral sulcus typical for diplodocines. However, some apatosaur specimens also have weak flanges, but no continuous ventral sulcus marking the ventral surface (BYU 1252-18531, NSMT-PV 20375 and UW 15556).

C180: Mid- and posterior cervical prezygapophyses, articular surfaces: flat (0); strongly convex transversely (1) (U95, U98-89; Fig. 54).

C181: Mid- and posterior cervical vertebrae, pre-epipophysis: absent (0); present (1) (Remes, 2007; Figs. 40 and 51).

Comments. The pre-epipophysis is herein defined as a rugose, horizontal ridge laterally below the prezygapophyseal facet, which connects with the prdl anteriorly.

C182: Mid- and posterior cervical vertebrae, spinoprezygapophyseal lamina, anterior end: remains vertical, with the free edge facing dorsally (0); is strongly inclined laterally (sometimes roofing a lateral fossa in the prezygapophyseal process (1) (T13-117; modified; Fig. 55).

Comments. At a first glance, it appears possible that this character is correlated with the occurrence of transversely convex prezygapophyseal facets. However, this is not the case, as can be seen in the several varying scores for these two characters.

C183: Mid- and posterior cervical neural arches, lateral fossae on the prezygapophysis process: absent (0); present (1) (Harris, 2006b; C12b-124; modified by T13-118; Figs. 51 and 55).

Comments. Where such a lateral fossa is present, it is dorsally roofed by a laterally tilted anterior end of the sprl. However, not all specimens with a laterally tilted lamina also bear these fossae, which justifies the use of two independent characters. The character was first used in a phylogenetic analysis by Carballido et al. (2012b), but was modified by Tschopp & Mateus (2013b) in order to include posterior cervical vertebrae as well.

C184: Mid- and posterior cervical vertebrae, prezygapophyseal centrodiapophyseal fossa: single cavity (0); subdivided into two cavities by a ridge (1); several accessory laminae subdivide the fossa into various smaller partitions (2) (Gilmore, 1936; U04b-2; modified; Figs. 38 and 40). Ordered.

Comments. A third state was added in order to be able to accurately code the holotype specimen of Barosaurus lentus (YPM 429), as well as a few other specimens. Two specimens coded as ‘0’ actually only preserve mid-cervical vertebrae (AMNH 7535, CM 3452, E Tschopp, pers. obs., 2011). It would thus be possible that more posterior elements of these cervical columns had subdivided prcdf. The character is treated as ordered, because an increase in lamination is thought to happen during ontogeny as well (Schwarz et al., 2007).

C185: Mid- and posterior cervical neural arches, centroprezygapophyseal lamina: single (0); dorsally divided, resulting in a lateral and medial lamina, the medial lamina being linked with interprezygapophyseal lamina and not with prezygapophysis (1); divided, resulting in presence of “true” divided centroprezygapophyseal lamina, dorsally connected to prezygapophysis (2) (U95; modified by C12b-127; Fig. 54).

Comments. Usually, taxa with “true” divided cprl also have a lamina connecting from the base of the cprl to the tprl.

C186: Mid- and posterior cervical transverse processes: posterior centrodiapophyseal lamina (pcdl) and postzygodiapophyseal laminae (podl) meet at base of transverse process (0); pcdl and podl do not meet anteriorly, postzygapophyseal centrodiapophyseal fossa extends onto posterior face of transverse process (1) (New; Fig. 56).

C187: Mid- and posterior cervical vertebrae, accessory horizontal lamina in center of spinodiapophyseal fossa, not connected with any surrounding laminae: absent (0); present (1) (New; Fig. 57).

Comments. This accessory lamina could be a vestigial version of the epipophyseal-prezygapophyseal lamina (sensu Wilson, 2012) or the accessory lamina connecting the podl with the sprl (as used herein, following Carballido et al., 2012b). However, because no connection exists with any surrounding lamina, this cannot be definitely confirmed in the cases included here. The use of an independent character is thus preferred. The lamina is generally situated in the center of the sdf.

C188: Mid- and posterior cervical vertebrae, posterior centrodiapophyseal lamina: is single (0); bifurcates towards its anterior end (1) (U04b-5; wording modified; Fig. 57).

Comments. Evidence from SMA 0011 shows that the presence of anteriorly bifurcated pcdl sometimes are a precursor of entirely double pcdl (see above). However, because in various specimens only bifurcated and not entirely double pcdl exist, the character was retained as independent from the one describing the single or double pcdl (see character 136).

C189: Mid- and posterior cervical vertebrae, centropostzygapophyseal lamina (cpol): single (0); divided, with medial part contacting interpostzygapophyseal lamina (1) (C12b-128; Fig. 56).

C190: Mid- and posterior cervical neural arches, interpostzygapophyseal lamina projects beyond the posterior margin of the neural arch (including the centropostzygapophyseal lamina), forming a prominent subrectangular projection in lateral view: absent (0); present (1) (D12-26; modified by M13-131; Fig. 40).

Comments. A reduced subrectangular projection is present in mid-cervical vertebrae of Supersaurus WDC DMJ-021. Generally, the development of this feature increases in more posterior elements (e.g., in Diplodocus carnegii CM 84; Hatcher, 1901). Supersaurus WDC DMJ-021 was thus scored as apomorphic, although it is not prominent in the preserved vertebrae. On the other hand, Apatosaurus louisae CM 3018, where only CV 13–15 bear weak projections, was coded as plesiomorphic.

C191: Mid- and posterior cervical vertebrae, postzygapophyseal centrodiapophyseal fossa and spinopostzygapophyseal fossa: entirely separated (0); connected by a large foramen (1) (New; Fig. 36).

Comments. The laminae in this area are very thin and might break easily. In fact, many specimens do show an opening here, but most of them also show broken margins around this opening, making it impossible to decide if the feature is genuine or not. Often, possible foramina are also closed with plaster or similar material during preparation, probably for stability reasons, and because the presence of such foramina has never been reported before. In fact, only SMA 0011 can be confidently scored as apomorphic to date.

C192: Posterior cervical vertebrae, Elongation Index (cervical centrum length, excluding condyle, divided by posterior centrum height): less than 2.0 (0); 2.0–2.6 (1); higher than 2.6 (2) (G86; M12-90, 91; modified; Table S23).

Comments. In vertebrae with inclined posterior edges of the anterior condyle, a vertical line is drawn through the posterior-most point of the posterior edge, and the horizontal distance from this vertical line to a second vertical line through the posterior-most extension of the centrum is measured and taken as centrum length in this case. In some cases, only measurements of the complete centrum length were available, and the EI for the centrum length without anterior ball was calculated based on the mean difference between EI with and without condyle. Singular ratios given in Table S23 have to be taken with care, as they differ considerably within posterior cervical centra (decreasing towards posterior). Ratios based only on anterior posterior cervical vertebrae thus have to be corrected to a lower ratio (e.g., in UW 15556, Table S23). A simple EI is preferred over an average EI (centrum length divided by the mean of posterior centrum height and width; Chure et al., 2010) because many specimens could not be measured directly and lack published measurements. Therefore, many OTUs included herein had to be scored based on figures. Given that the lateral view is often the only one provided, reasonable comparisons could only be made when using the simple version of the EI.

C193: Posterior cervical vertebrae, ventral keel: anteriorly placed (0); restricted to posterior portion of centrum (1) (New; Fig. 53).

Comments. Taxa without ventral ridges are scored as unknown. The posterior restriction of the keel was proposed as an autapomorphy of Dinheirosaurus lourinhanensis by Mannion et al. (2012).

C194: Posterior cervical prezygapophyses: terminate with or in front of articular ball of centrum (0); terminate well behind articular ball (1) (U04b-3; modified; Fig. 40).

Comments. The neural canal should be held horizontally, in order to accurately assess the expansion of the prezygapophysis.

C195: Posterior cervical vertebrae, prezygapophysis articular facet posterior margin: confluent with prezygapophyseal process (0); bordered posteriorly by conspicuous transverse sulcus (1) (T13-121; Figs. 39 and 55).

Comments. The distribution of this character is dubious, because it is difficult to observe in photographs and drawings. To date, only the holotype specimen of Kaatedocus siberi (SMA 0004) was reported to bear such a sulcus. The character in its present state thus does not contribute to the resolution of the tree. It was retained because more work on actual specimens has to be performed in order to confirm or discard this character as an unambiguous autapomorphy of K. siberi.

C196: Posterior cervical vertebrae, spinoprezygapophyseal lamina: continuous (0); developing an anterior projection (just beneath but independent from the spine summit) (1) (T13-124; Fig. 39).

Comments. Sometimes the spine summit projects anteriorly (in particular in dicraeosaurs), which is not what this character describes. Diplodocines often have an anterior projection below the summit, which forms the most anterior point of the spine.

C197: Posterior cervical vertebrae, accessory lateral lamina connecting postzygodiapophyseal and spinoprezygapophyseal laminae: absent (0); present (1) (G05-25; Fig. 36).

Comments. This lamina was termed epipophyseal-prezygapophyseal lamina by Wilson & Upchurch (2009), but there are different ways of how to unite the epipophysis with the prezygapophysis (Carballido et al., 2012b; Wilson, 2012). Therefore, the description of Carballido et al. (2012b) was preferred herein.

C198: Posterior cervical vertebrae, accessory, subvertical lamina in the postzygapophyseal centrodiapophyseal fossa, with free edge facing laterally: absent (0); present (1) (New; Fig. 56).

Comments. Two types of accessory laminae occur in the pocdf of certain sauropod taxa: (1) laterally facing, relatively broad laminae, which are mostly located posteriorly, marking the lateral wall of the neural canal, and (2) more distinct, posteriorly facing laminae connecting the pcdl and podl anteriorly, at the base of the transverse process. The present character describes the presence of the first type, and the second type is accounted for in character 199.

C199: Posterior cervical vertebrae, accessory, subvertical lamina in the postzygapophyseal centrodiapophyseal fossa, with free edge facing posteriorly: absent (0); present (1) (Gilmore, 1936; U04b-6; modified; Fig. 36).

Comments. This accessory lamina is the one character 95 of Mannion et al. (2012) codes for. Rarely, posteriorly facing accessory laminae appear as a parallel pair (e.g., SMA 0011; Fig. 36). Jobaria has posteriorly facing laminae in the posterior portion of the pocdf, connecting to the postzygapophyses. They are herein interpreted as lateral cpol, which are somewhat anteriorly shifted. Jobaria is thus scored as plesiomorphic in this character.

C200: Posterior cervical postzygapophyses: terminate at or beyond posterior edge of centrum (0); terminate in front of posterior edge (1) (U04b-4; modified by T13-129; Fig. 57).

C201: Posterior cervical neural arch, interpostzygapophyseal lamina (tpol): connects directly with roof of neural canal (0); vertical lamina connects tpol with neural canal roof (1) (New; Fig. 56).

Comments. Carballido & Sander (2014) termed this vertical lamina ‘single intrapostzygapophyseal lamina’ (stpol).

C202: Posterior cervical neural arches, epipophyses: transversely compressed (0); dorsoventrally compressed (1) (New; Fig. 42).

Comments. Two different morphologies of the epipophyses occur in diplodocids: (1) dorsoventrally compressed, usually forming a horizontal, rugose ridge above the postzygapophyseal facet, on the lateral side of the spol, and (2) transversely compressed, such that it is formed by a dorsal expansion of the posterior end of the spol, in some cases (e.g., Diplodocus carnegii CM 84) forming a rugose, vertical plate above the zygapophyseal facet, but never accompanied by a horizontal ridge. Taxa without epipophyses are scored as unknown.

C203: Posterior cervical neural arches, accessory spinal lamina: absent (0); present, running vertically just posterior to spinoprezygapophyseal lamina (1) (W11-98; Fig. 40).

Comments. This lamina could represent a reduced spdl. The presence of a distinct lamina is restricted to advanced diplodocines, but a reduced lamina is present in Spinophorosaurus as well (NMB-1699-R, E Tschopp, pers. obs., 2011).

C204: Posterior cervical neural spines, dorsoventrally elongate coel on lateral surface: absent (0); present (1) (M12-99; Fig. 53).

C205: Posterior cervical neural spines, horizontal, rugose ridge right below spine summit on lateral surface: absent (0); present, serves as distinct dorsal edge of the spinodiapophyseal fossa (1) (T13-127; Fig. 42).

Comments. The ridge is slightly curved in some specimens (e.g., SMA 0011). When absent (plesiomorphic state), the sdf fades dorsally.

C206: Posterior bifid, cervical neural spines, medial surface: marked by distinct, dorsoventral ridge from base to spine summit (0); smooth (1) (New; Fig. 58).

C207: Posterior cervical neural and/or anterior-most dorsal neural spines: vertical (0); anteriorly inclined (1) (R05-71).

Comments. See comments in character 169 for definition of inclined.

C208: Posterior cervical and anterior dorsal vertebrae, roughened lateral aspect of prezygodiapophyseal lamina: absent (0); present (1) (W11-102; Fig. 59).

Comments. The rugose area in the derived taxa lies ventrolateral to the pre-epipophysis, when present.

C209: Posterior cervical and anterior dorsal vertebrae, prespinal lamina: absent (0), present (1) (S97-14, modified; Fig. 60).

Comments. The presence of a prespinal lamina does not imply the presence of a median tubercle or vice versa. However, a dorsally expanded prespinal lamina can form a median tubercle (see below). In anterior dorsal vertebrae of Diplodocus carnegii CM 94, the median tubercle leans anteriorly, but no lamina connects it with the base of the notch between the metapophyses (E Tschopp, pers. obs., 2011).

C210: Posterior cervical and anterior dorsal bifid neural spines, median tubercle: absent (0); present (1) (McIntosh, 1990b; U95; Fig. 39).

Comments. The median tubercle can be either an independent structure in the trough between the metapophyses, or a dorsal projection of the prespinal lamina.

C211: Posterior cervical and anterior dorsal bifid neural spines, orientation: diverging (0); parallel to converging (1) (R05-74; modified; Fig. 60).

Comments. Some taxa have diverging neural spines, with only their summits approaching an almost parallel orientation (e.g., CM 11984 or USNM 10865). They are scored as plesiomorphic herein. The character was initially proposed including the rate of divergence (Rauhut et al., 2005). The character was divided because the dorsal portions of the metapophyses can be parallel, but still widely separated from each other, as is the case in Camarasaurus.

C212: Posterior cervical and anterior dorsal bifid neural spines, divergence: wide (0); narrow, distance between spine summits subequal to neural canal width (1) (R05-74; modified; Fig. 60).

Comments. This is the second part of the character proposed by Rauhut et al. (2005; see character 211).

C213: Posterior cervical, and anterior and mid-dorsal vertebrae, anterior projection of diapophysis laterally adjacent to prezygapophyseal facet: absent (0); present (1) (New; Fig. 58).

Comments. The projection described herein is not to be confused with the projection sometimes formed by the pre-epipophysis, which is posteriorly accompanied by a horizontal, rugose ridge.

C214: Cervical ribs, length: long, reaching posterior to posterior end of centrum (0); short, not reaching posterior end of centrum (1) (R93-12; modified; Fig. 51).

Comments. An additive binary version describing cervical rib length is preferred herein over the multistate character of Whitlock (2011a).

C215: Cervical ribs, length: overlapping several centra posterior (0); overlapping no more than the next cervical vertebra in sequence (1) (R93-12; modified; Fig. 41).

C216: Cervical ribs, position relative to centrum: not projecting far beneath centrum (0); projecting well beneath centrum, such that length of posterior process is subequal in length to fused diapophysis/tuberculum (1) (Wilson, 2002; W11-153; modified; Fig. 40).

Comments. Whitlock (2011a) included two characters describing the length of the ventral projection (from Wilson, 2002) and comparing the length of the posterior process with the length of the fused diapophysis/tuberculum. However, the length of the fused diapophysis and tuberculum depends on how far the cervical ribs project ventrally, and the length of the posterior process is accounted for in the characters defining cervical rib length. Wilson (2002) defined the ventral projection as strong when it leads to a vertebral height that exceeds its length. Such a ratio is also present in dicraeosaurids, but because of their highly elevated neural spines. The ventral projection of the cervical rib of dicraeosaurids is minimal as in all taxa other than apatosaurs. Therefore, the two characters of Wilson (2002) and Whitlock (2011a) are herein combined, in order to define ventral projection compared to the length of the posterior process of the cervical rib.

C217: Cervical ribs, posteriorly projecting spur on dorsolateral edge of posterior shaft: absent (0); present (1) (New; Fig. 49).

Comments. The spur was proposed as autapomorphic for Turiasaurus (Royo-Torres, Cobos & Alcalá, 2006), but it is also present in some apatosaurs and Dicraeosaurus (E Tschopp, pers. obs., 2011; E Tschopp, pers. obs., 2012).

C218: Anterior and mid-cervical ribs, tuberculum in lateral view: is directed nearly vertically (0); is directed upwards and backwards (1) (U04b-12; modified; Fig. 50).

Comments. The orientation of the tuberculum tends to become more vertical in more posterior elements. Some apatosaurs scored as plesiomorphic here actually do not have any anterior cervical vertebrae preserved, which means that they could still have inclined tubercula in the anterior elements. However, because others have distinctly inclined tubercula in mid-cervical ribs as well, a differential coding is still justifiable. Taxa that do not preserve cervical ribs were coded based on the relative positions of diapophysis and parapophysis.

C219: Posterior cervical ribs, anterior process: present (0); absent (1) (U04b-9; modified; Fig. 40).

C220: Posterior cervical ribs, anterior process: distinct, much longer anteroposteriorly than high dorsoventrally (0); reduced to a short bump-like process or absent (1) (New; Fig. 61).

Comments. The last two characters serve as additive binary characters describing the reduction of the anterior process in apatosaurs in general and its complete absence in some apatosaur specimens (e.g., CM 3018; Gilmore, 1936; Wedel & Sanders, 2002).

C221: Posterior cervical ribs, anterior process: rounded in lateral view (0); has an acute pointed tip in lateral view (1) (U04b-10; modified; Fig. 61).

Comments. The anterior processes of cervical ribs can be rounded in dorsal view, but dorsoventrally compressed (as in SMA 0011, see above). Therefore, they are still pointed in lateral view.

C222: Posterior cervical ribs, rounded sub-triangular process in lateral view, posteroventral to tuberculum: absent (0); present (1) (Wedel & Sanders, 2002; U04b-11; wording modified; Fig. 61).

Comments. Upchurch, Tomida & Barrett (2004) scored the holotypic cervical vertebra of Apatosaurus laticollis YPM 1861 as plesiomorphic. However, as Wedel & Sanders (2002) showed, a process is clearly present in this specimen.

C223: Posterior cervical rib shafts: nearly straight and directed backward and a little upwards (0); initially directed in same direction but turn to run a little downwards toward distal tip (1) (U04b-13; Fig. 61).

Dorsal vertebrae

C224: Number of dorsal vertebrae: 13 or more (0); 12 (1); 10 (2); 9 (3) (McIntosh, 1990b; R93-14; modified; Table S24).

Comments. Amargasaurus was initially described to have 9 dorsal vertebrae (Salgado & Bonaparte, 1991), but the putative first dorsal has the parapophysis positioned dorsally to the pleurocoel, which is highly unusual in sauropods (Carballido et al., 2012a). Generally, this position marks the second or third dorsal vertebrae, which means that there would be at least ten dorsal elements, which was the coding used by Mannion et al. (2012). Herein, a coding as unknown is preferred, following Carballido et al. (2012b).

C225: Dorsal centrum length (excluding articular ‘ball’), remains approximately the same along the sequence (0); shortens from anterior to posterior dorsal vertebrae (1) (M12-106; Table S25).

Comments. The exclusion of the articular ball for measuring centrum length for this character is crucial, because anterior dorsal vertebrae often have considerably larger anterior condyles than posterior elements. In taxa lacking measurements or good figures to compare between anterior and posterior elements, scores of Mannion et al. (2012) were used (e.g., Omeisaurus).

C226: Dorsal vertebrae, opisthocoely (including a prominent anterior articular ‘ball’) disappears: between DV2 and DV3 (0); between DV3 and DV4 or more posteriorly (1) (Holland, 1915a; Gilmore, 1936; U04b-15; Table S26).

Comments. The definition of ‘prominent anterior ball’ is somewhat ambiguous. However, a new definition is not given here, because the character is interpreted to describe a significant change within the same vertebral column. These changes can be of different absolute size if one compares between specimens, but are relatively obvious within the same individual. The decrease is thus relative to its development in more anterior elements, but can be low in an absolute sense.

C227: Dorsal pneumatopores (pleurocoels): present (0); absent (1) (G86; McIntosh, 1990b; U95; polarity reversed; Fig. 62).

Comments. The dorsal centra of all included sauropod taxa have pleurocoel-like depressions on their lateral side, but in some taxa they do not bear a foramen.

C228: Dorsal centra, pneumatic structures: absent, dorsal centra with solid internal structure (0); present, dorsal centra with simple and big air spaces (1); present, dorsal centra with small and complex air spaces (2) (W02-77; modified by C12b-139; Fig. 37).

C229: Dorsal neural arches, paired, subdivided pneumatic chambers dorsolateral to neural canal: absent (0), present (1) (Sereno et al., 1999; W11-106; Fig. 63).

Comments. Paired pneumatic foramina occur in some diplodocids (e.g., UW 15556, YPM 1840), but they are not subdivided and are far less deep than in Nigersaurus or Demandasaurus. The latter are thus the only taxa with the apomorphic state.

C230: Dorsal transverse processes, orientation: horizontal or only slightly inclined dorsally (0); more than 30° inclined dorsally from the horizontal (1) (Y93-58; modified by U98-102; Fig. 63).

Comments. The angle of the transverse processes is easily affected by diagenetic distortion, as can be seen in DV 3 of Suuwassea ANS 21122, which most probably would actually have horizontal transverse processes.

C231: Dorsal vertebrae, single (not bifid) neural spines, spinoprezygapophyseal laminae: separate along entire length (0); joined distally, forming single prespinal lamina (1) (U95; modified by W11-107; Fig. 64).

Comments. In some taxa (e.g., Losillasaurus or Camarasaurus), the sprl unite dorsally with the prsl, but remain separate up to that point. Here, only taxa where the prsl is formed by the junction of the two sprl are scored as apomorphic.

C232: Dorsal vertebrae, spinodiapophyseal webbing: laminae follow curvature of neural spine and diapophysis in anterior view (0); laminae ‘festooned’ from spine, dorsal margin does not closely follow shape of neural spine and diapophysis (1) (S07-43; Fig. 65).

C233: Dorsal vertebrae with single neural spines, middle single fossa projected through midline of neural spine: present (0); absent (1) (C12b-144; Fig. 64).

Comments. The fossa described herein is a distinctly confined area within the sprf, restricted to the anterior edge of the neural spine process.

C234: Dorsal (single) neural spines, postspinal lamina, dorsal end: flat to convex transversely (0); concave transversely (1) (New; Fig. 66).

C235: Dorsal vertebrae, transition from bifid to single neural spines: gradual (0); abrupt (1) (New).

Comments. Gradual transitions go from deeply bifid, to shallowly bifid, to notched, to unsplit, as defined by Wedel & Taylor (2013). If one of the intermediate states is lacking, the taxon is scored as derived. Obviously, only specimens with articulated dorsal vertebrae can be scored for this character. Taxa without spine bifurcation are scored as unknown.

C236: Dorsal neural arches, hyposphene-hypantrum articulations: present (0); absent (1) (G86; S97-25; Table S27).

C237: Dorsal vertebrae, hyposphene first appears: on DV3 (0); on DV4 or more posteriorly (1) (U04b-19; modified; Table S27).

Comments. Both in Apatosaurus and Camarasaurus there are differences in the appearance of the hyposphene (Ikejiri, 2004; Upchurch, Tomida & Barrett, 2004). Because the type species, C. supremus, appears to show the plesiomorphic state, the genus was scored as such as well. Ikejiri (2004) suggests that the development of the hyposphene might depend on ontogeny, based on observations in the juvenile specimen CM 11338. However, the latter specimen is articulated and the region with the hyposphene is obliterated, such that its presence or absence is difficult to assess (McIntosh et al., 1996a).

C238: Dorsal vertebrae, single vertical lamina supporting the hyposphene from below: absent (0); present (1) (Gilmore, 1936; U04b-20; modified; Fig. 63).

Comments. The original character description (Upchurch, Tomida & Barrett, 2004) interfered with the character proposed by Wilson (2002) distinguishing between single and double cpol in mid- and posterior dorsal vertebrae (see character 261). The character of Upchurch, Tomida & Barrett (2004) was thus simplified, and polarity was reversed due to the differential taxon sampling. The lamina described herein corresponds to the stpol (Carballido & Sander, 2014). Taxa without hyposphene are scored as unknown.

C239: Dorsal vertebrae 1 and 2, centrum length: DV 1 > DV 2 (0); DV 2 > DV 1 (1) (U04b-14; modified; Table S28).

Comments. The character was originally defined implying that either DV 1 or 2 were the longest in the series (Upchurch, Tomida & Barrett, 2004), which is not always the case (see Table S28).

C240: First dorsal vertebrae, pleurocoel location: occupy the anterior and middle part of the centrum (0); occupy the posterior part of the centrum (1) (Holland, 1915a; Gilmore, 1936; U04b-17; modified; Fig. 59).

Comments. The character was restricted to the first dorsal, as also in Apatosaurus louisae, for which this character was proposed as a species autapomorphy (Holland, 1915a; Gilmore, 1936; Upchurch, Tomida & Barrett, 2004). In this taxon, DV 2 and 3 already have a centrally placed pleurocoel (CM 3018, E Tschopp, pers. obs., 2011).

C241: Anterior dorsal vertebrae, pleurocoels in first few centra: become larger along the series (0); become smaller (1) (Gilmore, 1936; U04b-16; wording modified; Table S29).

Comments. Taxa without dorsal pleurocoels are scored as unknown.

C242: Anterior dorsal vertebrae, ventral keel: absent (0); present (1) (M12-110; Fig. 67).

C243: Anterior dorsal transverse process position: high, considerably above dorsal edge of posterior cotyle (0); low, ventral edge about level to dorsal edge of posterior cotyle (1) (Gilmore, 1936; Fig. 68).

Comments. The differing dorsoventral extension of the transverse processes in the anterior-most dorsal vertebrae was proposed as character to distinguish Apatosaurus louisae CM 3018 from the supposed Apatosaurus excelsus UW 15556 (Gilmore, 1936). It is here applied for the first time in a phylogenetic analysis. In most taxa, position of the transverse process rises considerably dorsally in the first few dorsal vertebrae. Therefore, this description applies best for the first element in the series.

C244: Anterior, bifid dorsal vertebrae, base of notch between metapophyses: wide and rounded (0); narrow, V-shaped (1) (Gilmore, 1936; Fig. 68).

Comments. As observed in Apatosaurus, Camarasaurus also appears to show intrageneric variation: C. lewisi has narrow troughs throughout its bifurcated presacral vertebrae, whereas other Camarasaurus species have wide bases (Jensen, 1988; McIntosh et al., 1996b). Herein, Camarasaurus was scored as plesiomorphic, scoring the type species C. supremus.

C245: Anterior dorsal, bifid neural spines, medial surface: gently rounded transversely (0); subtriangular (1) (New; Fig. 60).

Comments. Some diplodocid specimens bear a dorsoventral ridge on the medial surface of the anterior dorsal neural spines, similar to the ridge present in some diplodocid posterior cervical neural spines. The ridge results in a subtriangular shape of the medial surface.

C246: Dorsal vertebra 3, parapophysis: lies at the top of the centrum (0); lies mid-way between the top of the centrum and the level of the prezygapophyses (1) (Gilmore, 1936; U04b-18; modified; Fig. 62).

C247: Anterior and mid-dorsal centra, pleurocoels: situated entirely on centrum (0); invade neural arch pedicels (1) (Holland, 1915a; Fig. 69).

Comments. Holland (1915a) proposed this morphology as diagnostic for Apatosaurus louisae. It is included in a phylogenetic analysis for the first time. Taxa without dorsal pleurocoels are scored as unknown.

C248: Anterior and mid-dorsal neural arch, hyposphene shape: rhomboid (0); laminar (1) (New; Table S27).

Comments. Hyposphene shape can change considerably from front to back, as is seen in specimens of Camarasaurus (Osborn & Mook, 1921; McIntosh et al., 1996b). In the present analysis, two different characters thus code for the anterior and mid-dorsal vertebrae, as well as for the posterior elements, which are often less developed (see character 276). See Fig. 63 for an example of a laminar hyposphene.

C249: Mid-dorsal neural arches, height above postzygapophyses (neural spine) to height below (pedicel): 2.1 or greater (0); <2.1 (1) (W11-114; modified; Table S30).

Comments. Pedicel height is measured from the neural canal floor to the ventral-most point of the postzygapophyseal facets, neural spine height from there to the spine top. Both measurements are taken vertically, ignoring spine inclination. The ratio changes considerably between mid- and posterior dorsal vertebrae, therefore the original character of Whitlock (2011a) was divided in two (see character 272). Furthermore, a numerical boundary was introduced.

C250: Mid-dorsal neural spines, form: single, bifid form (if present) does not extend past second or third dorsal (0); bifid, inclusive of at least fifth dorsal vertebrae (1) (W11-108; Table S31).

Comments. Notched and unsplit neural spines (sensu Wedel & Taylor, 2013) are counted as single; shallowly and deeply bifurcated spines as bifid. An additional character is used to account for the notched spines. The taxon scores are thus slightly different from the ones in Whitlock (2011a).

C251: Mid-dorsal neural spines, oblique accessory lamina connecting postspinal lamina with spinopostzygapophyseal lamina: absent (0); present (1) (New; Fig. 69).

Comments. In Supersaurus and Dinheirosaurus, this accessory lamina extends posterodorsally-anteroventrally from near the dorsal end of the posl to the junction of the spol with the spdl.

C252: Mid- and posterior dorsal vertebrae, lateral pleurocoels present in centra: absent (0); present (1) (G86; McIntosh, 1990b; U95; modified by W11-111).

C253: Mid- and posterior dorsal vertebrae, vertically oriented rod-like struts divide the lateral pneumatic foramina: absent (0); present (1) (M12-115; Fig. 69).

Comments. Mannion et al. (2012) proposed the presence of such a strut as a synapomorphy for the clade uniting Supersaurus and Dinheirosaurus. However, similar struts occur in some apatosaurs. The pleurocoel is often not completely liberated from matrix during preparation, potentially obscuring the presence or absence of this structure.

C254: Mid- and posterior dorsal vertebrae, height of neural arch below postzygapophyses (pedicel) divided by posterior cotyle height: <0.8 (0); 0.8 or greater (1) (G05-36; modified; Table S32).

Comments. Neural arch height is measured from the neural canal floor to where the postzygapophyseal facets meet medially, above the hyposphene, where present.

C255: Mid- and posterior dorsal neural arches, prezygoparapophyseal lamina: present (0); absent (1) (W02-97; Fig. 70).

C256: Mid- and posterior dorsal parapophyses, location: above centrum, posterior to anterior edge of centrum (0); straight above anterior edge of centrum, or anteriorly displaced (1) (New; Figs. 69 and 70).

Comments. The anterior edge of the centrum corresponds to the rim of the anterior condyle in opisthocoelous elements. In some taxa, the position of the parapophysis changes from front to back. These taxa are scored for the majority of the elements in the series (e.g., Haplocanthosaurus, where DV 10 has a posteriorly placed parapophysis, but the majority of the mid- and posterior dorsal vertebrae have anteriorly displaced parapophyses; Hatcher, 1903).

C257: Mid- and posterior dorsal neural arches, anterior centroparapophyseal lamina: absent (0); present (1) (U04a-133; modified; Fig. 70).

Comments. The character was herein adapted to restrict the positions to mid- and posterior caudal vertebrae, instead of including all dorsal vertebrae as in Upchurch, Barrett & Dodson (2004).

C258: Mid- and posterior dorsal neural arches, posterior centroparapophyseal lamina: absent (0); present as single lamina (1); present, double (2) (S97-22; modified after M13-148, based on D12-36; Figs. 70 and 71). Ordered.

Comments. In taxa, where the pcpl is double, the more dorsal branch often connects to the pcdl. Mannion et al. (2013) defined the third state as ‘two parallel laminae,’ but in certain specimens (e.g., Diplodocus carnegii CM 84), the dorsal branch becomes more horizontal (Hatcher, 1901). Mannion et al. (2013) based their character modification on character 36 of D’Emic (2012), which cites the occurrence of a single versus a double posterior centrodiapophyseal lamina (pcdl). However, this character should have referred to the posterior centroparapophyseal lamina (pcpl) rather than the pcdl (M D’Emic, pers. comm., 2015). Among the apomorphic features, D’Emic (2012) listed this character correctly as a double posterior centroparapophyseal lamina twice, referring to character 36 (D’Emic, 2012: appendices 3 and 4). Thus, character 36 of D’Emic (2012) is the same character as 148 of Mannion et al. (2013), and is included and slightly modified in our analysis. The character is treated as ordered, because it codes for both presence/absence and morphology.

C259: Mid- and posterior dorsal vertebrae, accessory laminae in region between posterior centrodiapophyseal lamina and posterior centroparapophyseal lamina: absent (0); present (1) (M12-116; Fig. 71).

Comments. This character is somewhat ambiguous. Some of these accessory laminae might actually represent dorsal branches of the pcpl (see character 258) or dislocated ppdl. Here, only laminae not directly connecting to any specifying landmark (see Wilson, 1999) are considered accessory. More studies are needed to see if these are homologous to the above mentioned laminae.

C260: Mid- and posterior dorsal vertebrae, accessory lamina linking hyposphene with base of posterior centrodiapophyseal lamina: absent (0); present (1) (New; Figs. 69 and 71).

Comments. The presence of such an accessory lamina was proposed as autapomorphic for Dinheirosaurus (Bonaparte & Mateus, 1999; Mannion et al., 2012), but is herein interpreted to occur in other diplodocids as well. The accessory lamina can easily be confused with the lateral branch of the cpol, but the latter connects directly with the postzygapophyseal facet and not with the hyposphene. The accessory lamina described herein is thus situated between the two branches of the cpol.

C261: Mid- and posterior dorsal neural arches, centropostzygapophyseal lamina: single (0); divided, lateral branch connecting to posterior centrodiapophyseal lamina (1) (W02-95; wording modified; Fig. 70).

Comments. The lateral branch is often only visible in lateral view.

C262: Mid- and posterior dorsal neural arches, infradiapophyseal pneumatopore between anterior and posterior centrodiapophyseal laminae: absent (0); present (1) (W02-103; Fig. 71).

Comments. Even though the development of pneumatic structures has been shown to depend on the ontogenetic stage (Wedel, 2003; Schwarz et al., 2007), the early juvenile brachiosaur SMA 0009 already has this pneumatopore.

C263: Mid- and posterior dorsal transverse processes, length: short (0); long (projecting <1.3 times posterior cotyle width) (1) (C12b-153; modified; Table S33).

Comments. The length of a single transverse process is compared to the maximum width of the posterior cotyle. Transverse process length is measured in a horizontal plane. Measurements taken from figures in posterior view generally underestimate the ratio, which has to be accounted for when scoring the taxa. In the case of Brachiosaurus altithorax FMNH P25107, true ratios based on the measurements by Riggs (1904) are about 120% of the ratios taken from published figures (Taylor, 2009), whereas in Apatosaurus NSMT-PV 20375 or Diplodocus CM 84, they are only 103% higher. This percentage depends on the relative position of the transverse processes above the centrum. Ratios generally decrease from anterior to posterior dorsal vertebrae. Taxa or specimens that preserve only posterior elements (e.g., Amphicoelias altus AMNH 5764) should thus have higher actual ratios than shown in Table S33.

C264: Mid- and posterior dorsal transverse processes, dorsal edge: straight, or curving downwards at distal end (0); developing a distinct dorsal bump or spur (1) (New; Fig. 63).

Comments. Spurs are usually situated at the distal tip, whereas bumps are located more medially.

C265: Mid- and posterior dorsal neural spines, anteroposterior width: approximately constant along height of spine, with subparallel anterior and posterior margins (0); narrows dorsally to form triangular shape in lateral view, with base being approximately twice the width of dorsal tip (1) (Taylor, 2009; M13-159; modified; Fig. 71).

Comments. Mannion et al. (2013) were the first to include this character in a phylogenetic analysis, based on observations by Taylor (2009), and encompassing the entire dorsal column. Herein, we restricted the character to mid- and posterior dorsal neural spines.

C266: Middle and posterior dorsal neural spines, breadth at summit: much narrower (0); equal to or broader (1) transversely than anteroposteriorly (W02-92; modified).

Comments. Neural spine width can change considerably from the spine bottom to the top. The original character was thus divided in two (see character 265).

C267: Mid- and posterior dorsal neural spines, triangular aliform processes: absent (0); present, do not project as far laterally as postzygapophyses (1); present, project at least as far laterally as postzygapophyses (2) (U98-116; modified after C12b-163; Figs. 63 and 64). Ordered.

C268: Posterior dorsal centra, total length/height of posterior articular surface: 1.0 or greater (0); short, <1.0 (1) (New; Table S34).

C269: Posterior dorsal centra, posterior articular surface width to height: 1.0 or less (0); >1.0 (1) (Gilmore, 1936; Table S34).

Comments. The boundary is set between 1.0 and 1.1 in the present study, because it was suggested by Gilmore (1936) to distinguish Apatosaurus louisae from A. ajax and A. excelsus.

C270: Posterior dorsal centra, articular face shape: amphicoelous (0); slightly opisthocoelous (1); strongly opisthocoelous (2) (Y93-40; wording modified by C12b-174; Fig. 70).

Comments. Slightly opisthocoelous means that the condyle is either ventrally or dorsally restricted, but still visible in lateral view. Strongly opisthocoelous vertebrae have anterior balls that reach from the dorsal to the ventral edge of the centrum. In Apatosaurus ajax YPM 1860, no anterior articulation surface of a posterior dorsal vertebrae is observable, but the posterior articulation surface of a posterior element has a small, but distinct fossa marking its upper half. This indicates a slightly opisthocoelous centrum in the following element.

C271: Posterior dorsal vertebrae, pleurocoel shape: oval to circular (0); subtriangular with apex dorsally (1) (New; Fig. 71).

Comments. Taxa without dorsal pleurocoels are scored as unknown.

C272: Posterior dorsal neural arches, height above postzygapophyses (neural spine) to height below (pedicel): <3.1 (0); 3.1 or greater (1) (W11-114; modified; Table S30).

Comments. See character 249.

C273: Posterior dorsal neural arches, parapophyseal centrodiapophyseal fossa: ventrally open, relatively shallow (0); deep, triangular (1) (G05-41; Fig. 71).

Comments. The apomorphic state is applied to specimens with the pcpl connecting to the pcdl or acdl, thus creating a ventrally closed, triangular fossa between them and the ppdl or prdl. In plesiomorphic taxa, the pcpl fades out posteroventrally or connects to the centrum anterior to the ventral end of the pcdl.

C274: Posterior dorsal vertebrae, spinoprezygapophyseal lamina: absent or greatly reduced (0); present (1) (U07-131; modified; Fig. 72).

Comments. Reduced sprl fade out anteroventrally and/or join the prsl at a very ventral level.

C275: Posterior dorsal postzygapophyses: almost horizontal, such that the two articular facets include a wide angle (0); articular facets oblique, including an almost 90° angle (1) (New; Fig. 66).

Comments. Some diplodocine taxa have curved facets. These are interpreted as horizontal because their lateral halfs are oriented horizontally.

C276: Posterior dorsal vertebrae, hyposphene-hypantrum system: well developed, rhomboid shape up to last element (0); weakly developed, mainly as a laminar articulation (1) (C12b-152; modified; Fig. 63; Table S27).

Comments. Taxa without hyposphenes are scored as unknown.

C277: Posterior dorsal neural arches, spinopostzygapophyseal laminae: single (0); divided near postzygapophyses (1) (W02-100; Fig. 63).

Comments. The spol can bifurcate in two ways in different taxa: rebbachisaurids have ventrally forked laminae, whereas in some diplodocids the spol bifurcates dorsally, creating a medial and a lateral branch. The presence of a medial spol is accounted for in character 278, the present one describes the ventral bifurcation.

C278: Posterior dorsal vertebrae, medial spinopostzygapophyseal lamina: absent (0); present and forms part of median posterior lamina (1) (C12b-172; Fig. 66).

Comments. The mspol can either be connected with the lspol ventrally or they can remain separated.

C279: Posterior dorsal vertebrae, base of neural spines just above transverse processes: longer than wide (0); subequal in width and length (1) (New).

Comments. This is the second character about spine width to length, inspired by a character from Wilson (2002) (see character 266).

C280: Posterior dorsal neural spines, orientation at its base: vertical (0); anteriorly inclined (1) (New; Fig. 70).

Comments. Anterior inclination can be restricted to the very base of the neural spine, as is the case in Apatosaurus louisae CM 3018 (Fig. 70A). The best indication for the inclination is the prsl in lateral view.

C281: Posterior dorsal neural spines, midline cleft along the dorsal surface: absent (0); present (1) (M12-121; modified; Fig. 65; Table S31).

Comments. The midline cleft described herein corresponds to the notched spines of Wedel & Taylor (2013). Not all posterior dorsal spines have to be notched in order to be scored as apomorphic.

C282: Posterior dorsal and/or sacral neural spines (not including arch), height: less than 2 times centrum length (0); 2–3 times centrum length (1); more than 3 times centrum length (2) (M12-123; modified; Table S35). Ordered.

Comments. Neural spine height is measured from the top of the postzygapophyses to the highest point of the spine, vertically. Centrum length does not include the anterior ball. The original version (Mannion et al., 2012) was restricted here to posterior dorsal and sacral vertebrae only, because mid-dorsal elements of diplodocids considerably lower the mean ratio in some cases (Table S35). Also, state boundaries are adapted.

C283: Dorsal ribs, rib head: area between capitulum and tuberculum flat (0); oblique ridge present that connects medial and lateral edge at the base of the rib head (1) (New; Fig. 73).

Comments. The ridge marks the posterior surface of the rib head of advanced diplodocines.

C284: Dorsal ribs, proximal pneumatopores: absent (0); present (1) (W02-141; Fig. 73).

Comments. In some taxa, only one rib of the entire series bears a pneumatopore. However, the ability to develop pneumatized ribs appears to be restricted to certain diplodocid groups, therefore the character was included in this analysis.

C285: Mid-dorsal ribs, orientation of tuberculum: spreading outside from rib shaft (0); following straight direction of rib shaft (1); following medial bend of rib shaft (2) (G05-39; Fig. 73).

Sacral vertebrae

C286: Sacral vertebrae, number: 4 (0); 5 (1); 6 (2) (S97-2; modified; Table S36).

Comments. Some Camarasaurus specimens appear to have six sacral vertebrae, which is usually considered a synapomorphy of advanced titanosauriforms (Tidwell, Stadtman & Shaw, 2005). The addition of a sacral vertebra was suggested to be a sign of very old age (Tidwell, Stadtman & Shaw, 2005). The unusual six sacral vertebrae in the holotype of ‘Apatosaurus’ minimus AMNH 675 (Mook, 1917) might thus also be ontogenetic.

C287: Sacral vertebral centra, pleurocoels: absent (0); present (1) (U04a-165; wording modified).

C288: Sacral rib III, ventral surface: smooth (0); with oblique ridge (1) (Mook, 1917; Fig. 74).

Comments. The presence of an oblique ridge was proposed as synapomorphy of Apatosaurus by Mook (1917), but later regarded as ambiguous and thus of little use to diagnose the genus (McIntosh, 1995). The presence of this ridge is herein used for the first time as a phylogenetic character, in order to test its utility. According to Mook (1917), the ridge marks the ventral face of sacral rib II. However, as shown in the holotype specimen of Brontosaurus amplus YPM 1981 (Ostrom & McIntosh, 1966), among others, the ridge actually lies on sacral rib III. Some Camarasaurus specimens bear oblique ridges on their sacral ribs (e.g., AMNH 690; Osborn, 1904), but not the genotype specimen AMNH 5761. In the present analysis, Camarasaurus was thus scored as plesiomorphic.

C289: Sacral neural spines, lateral side, towards summit: flat, with only spinodiapophyseal lamina (spdl) well-developed (0); with distinct horizontal accessory laminae that connect spdl to pre- and/or postspinal lamina (1) (New; Fig. 75).

C290: Sacral neural spines, lateral view, spinodiapophyseal lamina: reduced to absent, does not connect summit and diapophysis (0); present and distinct, connects spine summit with diapophysis (1) (New; Fig. 75).

C291: Sacral neural spines, lateral view, spinodiapophyseal laminae (spdl): remain vertical and thus parallel to each other (0); spdl of neighboring spines converge (1) (New; Fig. 75).

Comments. Diplodocinae develop a very distinct dorsal widening of the sacral spdl. Together with the inclination of the spines towards the central portion of the sacrum, this often leads to a fusion of these anteroposteriorly widened dorsal ends of the spdl.

Caudal vertebrae

C292: Caudal neural spines, elliptical depression between lateral spinal lamina and postspinal lamina on dorsolateral surface: absent (0); present (1) (S07-75; modified; Fig. 76).

Comments. Sereno et al. (2007) initially defined the character as follows: ‘elliptical depression between spinodiapophyseal lamina and postspinal lamina on lateral neural spine.’ However, the spinal lamina they were most probably referring to (herein called lateral spinal lamina) is usually the united spol and sprl (at least in diplodocids). The character description has thus been reworded in order to clarify this. Sereno et al. (2007) recovered the presence of such a depression as a synapomorphy of Nigersaurinae, but actually it is present in any taxon with transversely widened posl, and spol that either fuse with the spdl or the posl. Anterior caudal vertebrae of Diplodocus are a good example for this, although they were scored as plesiomorphic by Sereno et al. (2007). Taxa without spdl or posl are scored as unknown.

C293: Caudal neural spines with triangular lateral processes: absent (0); present (1) (S07-76; Fig. 77).

Comments. These processes correspond to the triangular lateral processes of dorsal neural spines, but do not appear to be correlated. They are restricted to anterior caudal vertebrae in the OTUs with the derived state included here, but because this is a simple presence–absence character, restriction to anterior caudal vertebrae is not necessary in the character definition.

C294: Posterior dorsal, sacral and anterior caudal neural spines, shape in anterior/posterior view: rectangular through most of length (0); ‘petal’ shaped, expanding transversely through 75% of its length and then tapering (1) (Calvo & Salgado, 1995; U98-117; Fig. 77).

Comments. Plesiomorphic caudal neural spines can still be transversely expanded at their ends. Also, taxa with gradually expanding neural spines that do not taper dorsally are herein scored as plesiomorphic, because without the tapering, the spines do not develop the ‘petal’ shape typical for rebbachisaurs and dicraeosaurs.

C295: First caudal centrum, articular face shape: flat (0); procoelous (1); opisthocoelous (2) (W02-116; modified).

Comments. The fourth state (biconvex) of Wilson (2002) was deleted because no OTU in this analysis has a biconvex first caudal vertebra. The probable brachiosaurid SMA 0009 and Demandasaurus have platycoel first caudal vertebrae (Torcida Fernández-Baldor et al., 2011; Carballido et al., 2012a), and are herein scored as opisthocoelous rather than flat.

C296: Anterior-most caudal centra, transverse cross-section: sub-circular with rounded ventral margin (0); ‘heart’-shaped with an acute ventral ridge (1) (Gilmore, 1936; U04b-22; wording modified; Fig. 78).

Comments. Taxa with ventral hollows in their anterior caudal centra are scored as plesiomorphic, because the presence of the ventral ridge is regarded as the crucial trait for which this character codes.

C297: Anterior-most caudal centra, pneumatic fossae: reduced to absent (0); large pleurocoels (1) (New; Fig. 76).

Comments. Some apatosaur specimens and Supersaurus have distinct pleurocoels in their anterior-most caudal centra, whereas in anterior centra (as defined in Table 3), pleurocoels are reduced to foramina in these taxa (see e.g., Riggs, 1903). The current character is thus added to the usual one coding for pleurocoels in anterior caudal vertebrae in general.

C298: Anterior-most caudal vertebrae, additional pneumatic fossa on posterodorsal corner of centrum: absent (0); present (1) (New; Fig. 76).

Comments. In lateral views, these additional pneumatic foramina are often obscured by the transverse process.

C299: Anterior-most caudal transverse processes, shape: triangular, tapering distally (0); wing-like (1) (McIntosh, 1990b; Y93-44; modified; Fig. 77).

Comments. A transverse process is herein interpreted as wing-like if it has a distinct shoulder, i.e., an angled bump on its dorsolateral edge.

C300: Anterior-most caudal vertebrae, transition from ‘fan’-shaped to ‘normal’ caudal ribs: between Cd 1 and 2 (0); Cd4 and Cd5 (1); Cd5 and Cd6 (2); Cd6 and Cd7 (3); Cd7 and Cd8 or more posteriorly (4) (U04b-23; modified; Table S37).

C301: Anterior-most caudal neural arches, accessory lamina connecting pre- and postzygapophyses: absent (0); present (1) (New; Fig. 76).

Comments. This accessory lamina usually connects the postzygapophysis with the sprl.

C302: Anterior-most caudal neural spine (not including arch), height: less than 1.5 times centrum height (0); 1.5 times centrum height or more (1) (Y93-59; modified after W11-126; Table S38).

Comments. Neural spine height is measured from the dorsal edge of the postzygapophyses to the spine top, vertically. Centrum height is measured at the posterior articular surface. Yu (1993) used the entire neural arch height for the ratio and formulated it as a multistate character, restricted to the first two caudal vertebrae. The ratio is herein adapted following Upchurch & Mannion (2009), but keeping the restriction to the anterior-most elements, instead of including all anterior caudal vertebrae as implemented by Upchurch & Mannion (2009).

C303: Anterior-most caudal neural spines, lateral spinal lamina: has the same anteroposterior width ventrally and dorsally (0); expands anteroposteriorly towards its distal end, and becomes rugose (1) (Upchurch, Barrett & Dodson, 2004; Fig. 76).

Comments. SMA 0087 appears to show the plesiomorphic state. However, due to the bad preservation of the bones, the true morphology of the lateral spinal lamina is difficult to assess, and it might actually turn out to be widened as in apatosaurines, once all of the material is prepared.

C304: Anterior caudal centra (excluding the first), articular surface shape: amphiplatyan or amphicoelous (0); procoelous/distoplatyan (1); slightly procoelous (2); procoelous (3) (McIntosh, 1990b; R93-17; modified after G09-52; Table S37).

Comments. The definition of “slightly procoelous” in this character is the same as for the “slightly opisthocoelous” in posterior dorsal centra (see character 270). In diplodocids, the centra change their shape in anterior to middle caudal vertebrae from slightly procoelous to procoelous/distoplatyan to amphicoelous/amphiplatyan. This change occurs more posteriorly in Diplodocus than in Apatosaurus, for example. Therefore, specimens of the former genus have to be scored as slightly procoelous for this character, whereas Apatosaurus specimens are scored as procoelous/distoplatyan. However, more detailed studies about this transition are needed in order to score this character appropriately, because the specimens used herein generally show some correlation (within Flagellicaudata) of the development of procoely and the presence of wing-like transverse processes, which also mark more caudal vertebrae in Diplodocus than in less derived Flagellicaudata.

C305: Anterior caudal centra, ventral surface: without irregularly placed foramina (0); irregular foramina present on some anterior caudal vertebrae (1) (W11-133; Fig. 78).

Comments. Foramina can also be present in anterior caudal vertebrae without concave ventral surfaces (see Suuwassea emilieae ANS 21122; Harris, 2006b).

C306: Anterior caudal centra, pneumatopores (pleurocoels): absent (0); present (1) (McIntosh, 1990b; Y93-32).

Comments. Small pneumatopores also mark the lateral surfaces of the centra in non-diplodocine sauropods (e.g., Lourinhasaurus alenquerensis MIGM specimen, E Tschopp, pers. obs., 2012). The development of the pneumatopores as foramina or deep coels is described in character 307.

C307: Anterior caudal centra, pneumatopores: restricted to foramina (0); large coels present (1) (T13-173; modified; Fig. 79).

Comments. This character only codes for the anterior caudal vertebrae, excluding the anterior-most elements with wing-like transverse processes. The presence of a large coel in the latter is coded for in character 297. Taxa without pneumatopores are scored as unknown.

C308: Anterior caudal centra, pneumatopores: disappear by caudal 15 (0); present until caudal 16 or more (1) (McIntosh, 2005; Table S37).

Comments. McIntosh (2005) recognized this as character distinguishing between Diplodocus and Barosaurus, but it is applied for the first time as a phylogenetic character.

C309: Anterior caudal centra, length: subequal amongst first 20 (0); more or less doubling over first 20 (1) (U98-133; modified; Table S39).

Comments. Lengths were compared between the shortest element among the first three, and the longest preserved vertebrae within Cd 17 and 22 (or if this part of the tail is lacking, the longest element preserved). Taxa with a ratio of 1.5 or more are scored as derived.

C310: Anterior caudal vertebrae, concavo-convex zygapophyseal articulation: absent (0); present (1) (Wilson, 2002; W11-143; Fig. 77).

Comments. This character is similar to the one for cervical vertebrae, which describes the flat versus convex prezygapophyses of diplodocine cervical vertebrae. Wilson (2002) suggested that convex prezygapophyses and concave postzygapophyses are diagnostic for Diplodocus, but Whitlock (2011a) showed that Barosaurus also showed the derived state. During the current study, some apatosaur specimens also were observed to have the apomorphic condition (BYU 1252-18531, UW 15556, YPM 1860, YPM 1980, YPM 1981).

C311: Anterior caudal prezygapophyses, pre-epipophysis laterally below articular facet: absent (0); present (1) (New; Fig. 76).

Comments. A rugose horizontal ridge marks the lateral surface of the prezygapophysis of Diplodocus and very few other taxa, below the articular facet. The position corresponds to where the pre-epipophysis of cervical vertebrae is located and is thus termed equally here.

C312: Anterior caudal vertebrae, transverse processes: ventral surface directed laterally or slightly ventrally (0); directed dorsally (1) (W11-125; Fig. 77).

Comments. This character describes the orientation of the ventral edge of the transverse process in anterior or posterior view.

C313: Anterior caudal transverse processes, anterior diapophyseal laminae (acdl, prdl): reduced or absent (0); present, well defined (1) (W02-129; modified; see Fig. 79C, 315-1 for equivalent in posterior diapophyseal laminae).

Comments. The original character (Wilson, 2002) was split in two, because the development of the posterior centrodiapophyseal and the postzygodiapophyseal laminae differs between Apatosaurus and Diplodocus.

C314: Anterior caudal transverse processes, anterior centrodiapophyseal lamina, shape: single (0); divided (1) (W02-130; Fig. 76).

Comments. In contrast to dicraeosaurids or more basal diplodocoids, diplodocids have wing-like transverse processes, which are anteriorly supported by two independent laminae, which both originate on the centrum and thus classify as acdl (and the latter thus as divided or double). In advanced diplodocines, the lower of the two acdl is furthermore branching in two towards the transverse process.

C315: Anterior caudal transverse processes, posterior diapophyseal laminae (pcdl, podl): reduced or absent (0); present, well defined (1) (W02-129; modified; Fig. 79).

C316: Anterior caudal transverse processes, anteroposteriorly expanded lateral extremities: absent (0); present (1) (New; Fig. 78).

Comments. Backwards curving transverse processes are not necessarily anteroposteriorly expanded.

C317: Anterior caudal neural spines, maximum mediolateral width to anteroposterior length ratio: <1.0 (0); 1.0 or greater (1) (U98-141; modified by M13-32; Table S38).

Comments. The anteroposterior length of the spine is measured at the same level as the maximum mediolateral width, perpendicular to the inclination of the neural spine. The unusual plesiomorphic state of SMA 0087 within the apatosaur specimens might be due to diagenetic transverse compression.

C318: Anterior caudal neural spines, spinoprezygapophyseal lamina: absent, or present as small short ridges that rapidly fade out into the anterolateral margin of the spine (0); present, extending onto lateral aspect of neural spine (1) (W02-121; modified by M12-145; Fig. 76).

C319: Anterior caudal neural spines, spinopre- and spinopostzygapophyseal laminae contact: absent (0); present (1) (W02-122; Fig. 76).

C320: Anterior caudal neural arches, prespinal lamina: absent (0); present (1) (U95; Fig. 76).

Comments. Sauropod anterior caudal neural spines are generally rugose anteriorly and posteriorly, but only derived eusauropods develop distinct ridges or laminae.

C321: Anterior caudal neural spines, thickened anterior rim of prespinal lamina: absent (0); present (1) (G05-54; Fig. 76).

Comments. Specimens without prespinal lamina are scored as unknown.

C322: Anterior caudal neural spines, prespinal lamina or rugosity: terminate at or beneath dorsal margin of neural spine (0); project dorsally above neural spine (1) (W11-131; modified; see Fig. 79A, 324-1 for equivalent in postspinal lamina).

Comments. The original character (Whitlock, 2011a) was split in two, because in the anterior caudal vertebrae of Cetiosauriscus stewarti NHMUK R.3078 only the postspinal rugosity expands dorsally above the spine summit (Woodward, 1905). The character description was slightly changed in order to include taxa without distinct prsl.

C323: Anterior caudal neural arches, postspinal lamina: absent (0); present (1) (U95; Fig. 76).

Comments. See character 320. The two characters coding for the presence of pre- or postspinal laminae, are scored equally in the present analysis, as also in Wilson (2002), and might thus prove correlated in future. They were both retained herein as they distinguish between basal and derived non-neosauropod eusauropods and should thus have no influence on the relationships between ingroup diplodocids.

C324: Anterior caudal neural spines, postspinal lamina or rugosity: terminate at or beneath dorsal margin of neural spine (0); project dorsally above neural spine (1) (W11-131; modified; Fig. 79).

Comments. See character 322.

C325: Anterior caudal neural arches; hyposphenal ridge on posterior face of neural arch; present (0); absent (1) (U95; polarity reversed by M12-142; Fig. 80).

C326: Anterior caudal neural spines, shape: single (0); slightly bifurcate anteriorly (1) (W11-139; Fig. 77).

Comments. Anterior caudal neural spines can be bifid in two ways: anteroposteriorly and transversely. The former is coded for in characters 322 and 324, whereas the latter is described in the present character.

C327: Anterior caudal neural spines, maximum mediolateral width to minimum mediolateral width ratio: <2.0 (0); 2.0 or greater (1) (C08-239; Taylor, 2009; modified by M13-34; Table S38).

C328: Anterior caudal neural spines, lateral expansion at distal end: gradual, expanding through the last third of the neural spine (0); abrupt, restricted to distal fourth of neural spine (1) (New; Fig. 77).

C329: Anterior and mid-caudal vertebrae, ventrolateral ridges: absent (0); present (1) (U04a-183; Fig. 81).

Comments. Two horizontal ridges mark some diplodocid caudal centra: a lateral ridge and a ventrolateral ridge. Usually, only one of the two is present, which is interpreted as the lateral ridge, given its often rather dorsal position. The ventrolateral ridge as used herein does not describe the borders of the ventral longitudinal hollow of advanced diplodocines.

C330: Anterior and mid-caudal centra, ventral longitudinal hollow: absent (0); present (1) (McIntosh, 1990b; Y93-63; Fig. 78).

Comments. A ventral hollow is herein interpreted to be a longitudinal concavity occupying the entire ventral surface. Various taxa have very distinct posterior chevron facets with distinct ridges leading to them, thus creating a posteriorly concave ventral surface. However, these ridges often fade anteriorly. In some anterior diplodocine caudal centra, longitudinal struts subdivide the ventral hollow (e.g., Tornieria africana SMNS 12141a; Remes, 2006).

C331: Anterior- and mid-caudal vertebrae, ventral hollow depth: shallow, 10 mm or less (0); deep, >10 mm (1) (Curtice, 1996; Table S39).

Comments. Ventral hollow depth is used as a character distinguishing between Diplodocus and Barosaurus (Curtice, 1996; McIntosh, 2005). Curtice (1996) showed that a caudal centra with a ventral hollow depth of more than 10 mm can be confidently identified as Diplodocus, whereas shallower centra are typical for less derived diplodocines. Only very limited measurements were available, and the scoring was mainly based on descriptions and thus the subjective opinion of the respective authors. An interesting case is present in Tornieria, where the only preserved caudal vertebra of the holotype specimen (SMNS 12141a, Cd 2) has a deep ventral hollow, whereas the medial caudal vertebra of skeleton k (MB.R.2913) is only shallowly excavated (Remes, 2006). More detailed research is needed in order to sort this out.

C332: Mid-caudal vertebrae, ratio of centrum length to posterior height: <1,7 (0); 1,7 or greater (1) (Y93-45; modified; Table S39).

Comments. Usually, this character is included in analyses with its state boundary set at 2. In the present analysis, it was regarded more useful to put the boundary at 1.7, because some diplodocine taxa have ratios between 1.7 and 2. Generally, the ratio increases in more posterior elements, therefore specimens with only anterior mid-caudal vertebrae preserved (e.g., Diplodocus longus YPM 1920, see McIntosh & Carpenter, 1998) most probably would have higher ratios than indicated in the table.

C333: Mid-caudal vertebrae, lateral surface of centra: without longitudinal ridge at midheight (0); longitudinal ridge present, centra hexagonal in anterior/posterior view (1) (Upchurch & Martin, 2002; U04a-186; modified by W11-146; Fig. 81).

Comments. This ridge is not the same as the ventrolateral ridge described above, which is located below midheight.

C334: Mid-caudal centra, articular surface shape: cylindrical (0); quadrangular (1); trapezoidal (2); with flat ventral margin but rounded lateral edges (3) (W02-131; modified; Fig. 82).

Comments. The character was modified in order to be able to code for the various intermediate states between cylindrical, quadrangular, and triangular as described by earlier workers (Gallina & Apesteguía, 2005; Carballido et al., 2012b). Articular surfaces of a rather hexagonal shape are scored as cylindrical, because the hexagonal shape is created by the lateral ridge described in character 333.

C335: Mid-caudal centra ventral surface in lateral view: gently curved (0); greater portion straight, with expansions on both ends to form the chevron facets restricted to about last fourth of centrum length (1) (New; Fig. 81).

Comments. This description applies especially for anterior mid-caudal elements, more posterior vertebrae of derived specimens tend to develop a more gentle curvature. This can create problems in taxa preserving only posterior mid-caudal vertebrae. For instance, Tornieria specimen k is herein scored as plesiomorphic for this character. Caudal vertebrae from trench dd, however, indicate that Tornieria actually might show the derived state, but these have not been found in articulation, and because anatomical overlap with the referred specimens included herein is minimal, their attribution to the species should be regarded as doubtful.

C336: Mid-caudal posterior articular surface: concave (0); flat (1); convex (2) (New; Table S37).

C337: Mid-caudal neural arches: over the midpoint of the centrum with approximately subequal amounts of the centrum exposed at either end (0); on the anterior half of the centrum (1) (Huene, 1929; S97-15; Fig. 81).

Comments. For this character, the distance between pre- and postzygapophyses and their location above the vertebral centrum is regarded as reference. The pedicels can still be dislocated anteriorly in plesiomorphic taxa. This character is generally used as a titanosauriform synapomorphy (Salgado, Coria & Calvo, 1997; Wilson, 2002), but also is convergently present in some Diplodocus specimens (e.g., AMNH 223, or USNM 10865).

C338: Mid-caudal prezygapophyses: free (0); posteriorly interconnected by a transverse ridge, creating a triangular fossa together with the spinoprezygapophyseal laminae (1) (New; Fig. 83).

Comments. This transverse lamina marks the caudal vertebrae of Diplodocus longus YPM 1920 and might prove a valid autapomorphy for the species in the future.

C339: Mid-caudal prezygapophyses position: terminate at or behind anterior edge of centrum (0); project considerably beyond anterior edge of centrum (1) (New).

Comments. Only taxa where the prezygapophyses clearly overhang the centrum (i.e., recognizable without any need of measuring) are scored as derived.

C340: Mid-caudal neural spines, orientation: directed posteriorly (0); vertical (1) (McIntosh, 1990a; S97-10; modified; Fig. 81).

C341: Mid-caudal neural arch, anterior extreme of spine summit: smooth (0); developing a short anterior or anterodorsal projection, such that anterior edge of spine becomes slightly concave (1) (New; Fig. 84).

Comments. Such a spur might also be interpreted as pathologic or ontogenetic. However, its presence in the juvenile to subadult Apatosaurus (= Camarasaurus) grandis YPM 1901 suggests that ontogeny can probably be excluded as a cause. More studies are needed in order to confirm or refute pathology, in the meanwhile the character is kept in the analysis.

C342: Mid- and posterior caudal vertebral centra, articular surfaces: subequal in width and height or higher than wide (0); considerably wider than high (1) (S97-34; modified;