Introduction

The stegosaurid Hesperosaurus mjosi Carpenter, Miles & Cloward (2001), was named for a nearly complete skeleton from very low in the Morrison Formation in north-central Wyoming, USA. Carpenter and Galton (2001) reviewed the species of Stegosaurus named by Marsh and concluded that S. armatus could not be diagnosed at the species level with the available material, although two species (S. stenops, S. ungulates) were valid (S. longispinus was not considered). Maidment et al. (2008) challenged these conclusions, considering Hesperosaurus mjosi as a species of Stegosaurus (as S. mjosi), along with S. armatus from North America, and S. homheni (for Wuerhosaurus homheni) from China. They noted that Stegosaurus mjosi differed from S. armatus “as the former possesses the following primitive characters: atlas neural arches not fused to intercentrum in ontogenetically mature individuals, postzygapophyses not elevated significantly on posterior cervical vertebrae, neural arches of dorsal vertebrae not elongated above the neural canal, ossified epaxial tendons present, ribs distally expanded, caudal neural spines not bifurcated, enlargement of the distal end of the pubis. It also has the following autapomorphies: 11 dorsal vertebrae, fourth sacral vertebra not fused to sacrum, dorsal dermal plates longer anteroposteriorly than tall dorsoventrally.” (Maidment et al. 2008, p. 379).

The number of North American stegosaur genera and species raises the age old question of the species concept in palaeontology. The subject for dinosaurs was addressed in Carpenter and Currie (1990), and more recently by Paul (2008) for iguanodontids. The argument for species recognition in palaeontology is mostly philosophical because species recognition is limited to only a portion of the entire biological system, namely the hard parts, which among vertebrates is the skeleton (Raup and Stanley 1978; Donoghue 1985; see also Ereshefsky 2001). The application of the species concept in palaeontology, including dinosaurs, has long struggled with what criteria to use in taxonomy.

Early palaeontologists, especially Marsh and Cope, are frequently derided today for naming taxa based on scrappy, fragmentary material. For example, Marsh (1889) named Triceratops galeus for a nasal horn core collected from the Late Cretaceous near Denver, CO, USA. Leidy (1856) named four species of dinosaurs from among a hand-full of teeth surface collected by Hayden in the Late Cretaceous badlands along the Judith River of Montana (Leidy 1859). Then, of course, there is Iguanodon named by Mantell (1825) for teeth discovered by his wife. Some of the fragmentary materials named by these pioneers are borderline as to whether they are sufficiently unique for the name to be valid. For example, Cope (1876) named Monoclonius crassus for material collected along the Judith River. Dodson (1996) accepted the taxon as valid, whereas Sampson et al. (2008) did not.

It is easy today to ridicule these pioneers for naming taxa from scrappy material. However, we need to view their actions in the light of the times. The 1800s is when the natural world was just beginning to be studied scientifically. The large majority of the animals and plants we know today were given their scientific names during this time. The principles of zoological and plant nomenclature, roughed out previously by Linnaeus (1735), were being utilized in scientific descriptions. It is important to consider that many of these early palaeontological pioneers in dinosaurs also described modern organisms. Cope is especially noted for his work on reptiles, amphibians and fishes (Osborn 1931). Thus, these early workers had an understanding of modern organisms and the methodology for describing them. It was with this concept of organisms that they drew upon when faced with fragmentary specimens. These specimens became less fragmented once techniques were developed for hardening crumbly fossils and encasing larger ones in plaster of Paris jackets. Regardless, these pioneers were faced with fossils of organisms unlike anything living today. Clearly, the organisms were morphologically different and hence, to their thinking, warranted a distinct name; this same philosophy holds today in the naming of specimens as is evidenced by the “diagnosis.”

It was inevitable that the principles of taxonomy as applied to vertebrate fossils in the 1800s and early 1900s would be questioned as specimens accumulated. Much of this questioning came during the 1930s to 1940s, and again in the 1960s, by such workers as Simpson, Dobzhansky and Mayr, and it has been revived yet again with the expanded use of cladistics (e.g. Wheeler and Meier 2000). The problem, of course, is due to the limitations of only having hard parts, usually fragmented and incomplete, and which are often distorted due to crushing during sediment compaction. It is the unfortunate nature of vertebrate fossils, particularly dinosaurs, that does not allow the use of modern techniques to identify species, including molecular, colour or scale patterns of skin, behaviour, reproductive potential, vocalizations, etc. Therefore, all methodologies for the identification of dinosaur species must rely upon skeletal differences. But how much difference must two specimens show or have to warrant separation at the species level, let alone at the generic level? Unfortunately, there is no consensus to this conundrum (i.e. the species problem in palaeontology). As Raup and Stanley (1978) noted, sibling species differ very slightly from one another, whereas other species are so very different that their separation is obvious. But what level of “obvious” is acceptable? Barrett et al. (2005) diagnosed two small ornithischians from China, each with a single autapomorphy: Hexinlusaurus multidens (“marked concavity that extends over the lateral surface of the postorbital”) and Xiaosaurus dashanpensis (“proximally straight humerus”). Are these single autapomorphies sufficient? If so, then the situation is as Raup and Stanley (1978, p. 108) stated: “A species is a species if a competent specialist says it is.” This use of only autapomorphies to diagnose taxa may work some of the time, but other times it may not. For example, Farke (2007: 253) noted that “few characters are truly unique to the species T[orosaurus] latus, but the combination of characters is unique. Only the combination of features diagnoses the taxon” (emphasis by Farke 2007).

Lessons learned from species concept in dinosaur palaeontology

There is little question that Tyrannosaurus rex is a distinct taxon from Allosaurus fragilis because each can be identified based on very distinct morphological (typological) criteria. However, such a distinction is less apparent among several other different groups of dinosaurs. As Sampson et al. (2008) have noted, ceratopsid taxa are notoriously conservative in their postcrania, with almost all of the generic and specific criteria being based on differences in the skull architecture (Fig. 1); this point has been also noted by Dodson (1996) and Paul (2007, 2008), among others. Lambeosaurine hadrosaurs are also conservative in their postcrania, with the biggest differences also occurring in the skull (Fig. 2; Dodson 1975; Paul 2007, 2008); even the well known tyrannosaurids are remarkably similar (Fig. 3). The reverse may also be true where the skull is conservative, with the biggest differences in the postcrania. Such a situation is seen in Apatosaurus and Diplodocus (Fig. 4), which led to the skull of Apatosaurus being unrecognized for over 50 years (Fig. 5; Berman and McIntosh 1978).

Fig. 1
figure 1

Example of ceratopsians in which the similarities are greater than the differences. Two genera of long faced chasmosaurines Chasmosaurus belli (a) and Anchiceratops (b) versus two genera of short faced centrosaurines Styracosaurus ovatus (c) and Centrosaurus apertus (d). Compare with the two species of Centrosaurus: Centrosaurus apertus (d) and Centrosaurus nasicornis (e). The two Centrosaurus species are linked by the hook-like process over the frill fenestra, and are separated in part by the nasal horn. Courtesy of Gregory S. Paul. Scale bar 1 m

Full size image
Fig. 2
figure 2

Comparison of lambeosaurine hadrosaurs: Corythosaurus casuarius (a), Lambeosaurus lambei (b) and Hypacrosaurus stebingeri (c). Cranial differences are in the crest, whereas most postcranial differences are in the neural spine height. Other differences are minor. Courtesy of Gregory S. Paul. Scale bar 1 m

Full size image
Fig. 3
figure 3

Comparison of various adult tyrannosaurids: Gorgosaurus libratus (a), Albertosaurus sarcophagus (b), Tyrannosaurus rex (c) and Tarbosaurus bataar (d). Differences are minor at the generic level. Courtesy of Gregory S. Paul. Scale bars 1 m

Full size image
Fig. 4
figure 4

The skulls of Apatosaurus louisae (CM 11162) (a) and Diplodocus sp. (CM 11161) (b), showing how similar they are and why it took over 50 years for them to be separated. Scale bar 10 cm

Full size image
Fig. 5
figure 5

Although the skulls of Apatosaurus and Diplodocus are very similar (cf. Fig. 4), their postcranial skeletons differ considerably. Apatosaurus louisae (CM 3018 and CM 11162) in left lateral (a), dorsal (b), anterior (c, without cervicals), and posterior (d, without caudals). Diplodocus carnegie (CM 84 and CM 94) in left lateral (e), dorsal (f), anterior (g, without cervicals), and posterior (h, without caudals). Courtesy of Gregory S. Paul. Scale bars 1 m

Full size image

Species recognition in dinosaurs is even more difficult because one person’s species is another person’s genus. A case in point is the skull Marsh (1894) referred to Camptosaurus amplus, a name long accepted until Brill and Carpenter (2007) renamed the specimen Theiophytalia kerri. Besides morphological differences, they also noted that the skull actually came from the Lower Cretaceous Lytle Formation, not the Upper Jurassic Morrison Formation. This recognition of two genera from what had previously been one also occurred with the separation of Mantellisaurus from Iguanodon by Paul (2007; see Fig. 6).

Fig. 6
figure 6

Generic differences between iguanodontids is readily apparent among Ouranosaurus nigeriensis (a), Mantellisaurus atherfieldensis (b) and Iguanodon bernissartensis (c), yet only recently was Mantellisaurus separated from Iguanodon by Paul (2007). Courtesy of Gregory S. Paul. Scale bars 1 m

Full size image

More recently, parsimony analysis of specimens (rather than taxa) has been used to cluster them into possible species (e.g. Yates 2003; Upchurch et al. 2004). However, assigning genus or species names to clusters, especially to “sister clusters”, will always be philosophical, and hence the subject of debate. It may be a more useful tool for determining what taxon a particular specimen belongs to.

The species concept in Stegosaurus

The genus Stegosaurus was established by Marsh (1877) for very scrappy material collected approximately in the middle of the Morrison Formation (Turner and Peterson 1999). Named S. armatus, the morphology of the caudal vertebrae and large plate-like osteoderms define the genus, species and family. Unfortunately, in the context of considerably more material, including nearly complete skeletons, it is debatable whether the material is diagnostic at the species level (Carpenter and Galton 2001). A similar situation exists with Titanosaurus indicus as noted by Wilson and Upchurch (2003), who discussed the concept of ‘historical obsolescence’, that is characters once thought to be unique were later found to be more widely distributed through the discovery of new and better specimens.

In making their diagnosis of Stegosaurus, Maidment et al. (2008) did not base it on the holotype material, YPM 1850, but rather on a composite of various specimens that were assumed by them to be of a single species showing a wide range of non-ontogenetic variation throughout the skeleton. However, the variation cannot be replicated in other large samples of stegosaurids (e.g. Kentrosaurus aethiopicus), making the validity of the variations suspect. Furthermore, some characters stated by Maidment et al. (2008) to be variable in S. armatus are used to diagnose other taxa. For example, the enormously long and slender osteoderm spines of S. longispinus are “not considered a valid autapomorphy as dermal armour is likely to be extremely variable depending on age, size and the sex of the animal that bore it” (Maidment et al. 2008, p. 17). Yet Gigantospinus sichuanensis is diagnosed from a single autapomorphy: “possessing a parascapular spine that is at least twice the length of the scapula” (Maidment et al. 2008, p. 11). Based on the ontogenetic changes known to occur in the osteoderms of stegosaurs (Hayashi et al. 2009), and acknowledged by Maidment et al. (2008), dismissal of spine size leaves G. sichuanensis without any diagnostic features, and must by conditions advocated by Maidment et al. (2008) be considered a nomen dubium. Clearly characters dismissed in the diagnosis of one taxon, must also hold true for another.

In other instances, the statements used to justify synonymies are clearly in error. For example, Wuerhosaurus hemheni is placed in the genus Stegosaurus as a new combination S. hemheni because “the elements preserved are identical in most aspects to those of Stegosaurus armatus” (Maidment et al. 2008, p. 14). Except for the general features that unite all stegosaurs, this statement is misleading. The pelvis as illustrated has a very short preacetabular process, unlike the long process in Stegosaurus (compare Maidment et al. 2008: Figs. 8C and 2B). S. armatus lacks the deep fossa above the neural canal on the posterior side, which Maidment et al. used to diagnose S. hemheni from the Early Cretaceous of China (Dong 1973). The dorsal plate is totally unlike those of Stegosaurus armatus (as defined by Maidment et al. 2008). These differences argue for Wuerhosaurus hemheni being a distinct genus from the Late Jurassic Stegosaurus.

With regards to Stegosaurus armatus, I agree with the proposal of Galton (2010) to petition the International Zoological Commission to suppress Stegosaurus armatus and establish S. stenops, USNM 4934, as the name-bearing type. Doing this eliminates the assumed wide range of variation claimed by Maidment et al. (2008) and allows separation of S. stenops and S. ungulatus as advocated by Carpenter et al. (2001) and Carpenter and Galton (2001). This separation places Hesperosaurus msoji outside the morphological range seen by Stegosaurus as first advocated by Carpenter et al. (2001; see Figs. 7, 8). As Maidment et al. (2008, p. 13) noted, “Hesperosaurus has autapomorphies and retained plesiomorphies not seen in other stegosaurs [i.e., Stegosaurus].” Furthermore as Mateus et al. (2009) admit, Hesperosaurus is closely related to Dacentrurus, which only strengthens the separation of Hesperosaurus from Stegosaurus.

Fig. 7
figure 7

Comparisons of the skeletons of Stegosaurus stenops (USNM 4934) (a) and the holotype of Hesperosaurus mjosi (HMNH 001) (b). Based upon the criteria used to separate other dinosaurs, these two warrant separation at the genus level. a is courtesy of Gregory S. Paul. Scale bars 1 m

Full size image
Fig. 8
figure 8

Pelvis of Hesperosaurus mjosi holotype (HMNH 001) (a) and Stegosaurus stenops (DMNH 1483) (b). The preacetabular process is strongly divergent in H. mjosi, less so in S. stenops. Scale bars 10 cm

Full size image

The distinction of Hesperosaurus is heightened by the discovery of several additional specimens low in the Morrison Formation near the Howe Ranch on the west side of the Big Horn Mountains, Wyoming (first-hand obs. 2009; Siber and Möckli 2009). The most important of these are SMA 3074-FV01 (“Moritz”) and SMA 0018 (“Victoria”). Incorporating data from these specimens allows for a revised diagnosis of Hesperosaurus mjosi.

Revised diagnosis: Antorbital fenestra large versus very small in Stegosaurus stenops; maxilla short and deep, height equal to half of length, unlike S. stenops where height equals one-third of length; basisphenoid short, long in S. stenops; long-necked stegosaur with 13 cervicals versus 10 in S. stenops; short bodied with 13 dorsals versus 17 in S. stenops; neural arch of mid-dorsals low as in Huayangosaurus taibaii, not tall as in S. stenops; cervical ribs expanded distally, not expanded in S. stenops; tops of neural spines of proximal caudals rounded, not birfurcated as in S. stenops; dorsal edge of scapular blade with distinct step mid-length, no step (margins parallel) in S. stenops; preacetabular process strongly divergent, less so in S. stenops; distal end of prepubic process expanded vertically, unexpanded in S. stenops; distal end of postacetabular process swollen into knob-like structure, unswollen in S. stenops. Mid-dorsal to anterior caudal osteoderm plates vertically low and oval in shape, rather than tall and triangular as in S. stenops.

Conclusions

The palaeontological concept of species is problematic because it is based on osteological characters. The situation is made worse by the fragmentary nature of most specimens. It is no wonder then that there is no agreement as to whether a particular taxon is a valid species, let alone a valid genus. Certain dinosaur taxa currently accepted as valid genera (e.g. Lambeosaurus, Corythosaurus and Hypacrosaurus) could arguably be considered different species of the same genus (G. S. Paul, pers. comm. 2009). However, other taxa are so different that they warranted separation, as recently occurred by the separation of Mantellisaurus from Iguanodon by Paul (2008). The Stegosauria is osteologically a conservative group and this makes taxonomic distinctions problematic. Philosphically, it has been argued by some that Stegosaurus consists of two species in North America, S. mjosi and S. armatus, the latter of which is exceptionally variable. Such a wide range of variability is unique among the Dinosauria, except in waste-basket taxa (e.g. Megalosaurus), and for that reason is rejected. Finally, as Maidment et al. (2008, p. 379) acknowledged, Hesperosaurus mjosi has a unique “combination of autapomorphies and retained plesiomorphies”. I consider that this unique combination warrants recognition of the taxon as a separate genus, as originally proposed by Carpenter et al. (2001). Furthermore, Stegosaurus armatus should be considered a nomen dubium as advocated by Galton (2010) and replaced with S. stenops. Even with these recommended changes, I have no doubts that this is not the last word on the subject.