Scelidosaurus (Owen, 1863)

Norman, David B, 2021, Scelidosaurus harrisonii (Dinosauria: Ornithischia) from the Early Jurassic of Dorset, England: biology and phylogenetic relationships, Zoological Journal of the Linnean Society 191 (1), pp. 1-86 : 59-62

SCELIDOSAURUS

PREVIOUS ANALYSES AND DATA SOURCES

1. Wa l t e r C o o m b s ’ (1 9 7 1, 1 9 7 8a) r e v i e w s o f Ankylosauria clarified and greatly improved our understanding of their anatomy, and offered a logical basis (through the listing of sets of anatomical characteristics – effectively synapomorphies) that established a classification for the

As a measure of support for his stated position on the matter, Coombs was able to refer (using the note added in proof) to an account of Scelidosaurus ( Thulborn, 1977) , in which it was argued (incorrectly) that this dinosaur was ornithopodan rather than an ankylosaurian. The extent to which Coombs had been misled about the anatomy of Scelidosaurus reflects the want of a descriptive revision, rather than any shortcoming of his own.

Many of the anatomical characters listed by Coombs (1978a) in his review of the classification of Ankylosauria have been translated into characterstate formats in numerous subsequent numerical cladistic analyses.

2. Sereno (1986) summarized ornithischian relationships and provided synapomorphy lists to support the clade Thyreophora : Scutellosaurus + Scelidosaurus + Eurypoda).Five characters supported a more exclusive clade named Thyreophoroidea ( Scelidosaurus + Eurypoda – see Fig. 38 ):

(a) A sinuous dentary tooth row in lateral view. (b) A supraorbital bone forms the dorsal orbital

margin.

(c) Enlargement of the medial portion of the mandibular condyle.

(d) A basisphenoid that is much shorter than the basioccipital.

(e) Median palatal pterygovomerine keel.

3. Tumanova (1987) and Coombs & Maryańska (1990) did the same in supporting a cladogram summarizing ingroup relationships among ankylosaurs that reinforced the conclusions reached by Coombs (1978a). Sereno & Dong (1992) provided an updated description of the early stegosaur Huayangosaurus ( Dong et al., 1982) . In this account, a discussion of the anatomical features found in Huayangosaurus , as well as those shared by other known stegosaurs, was provided as a basis for justifying a synapomorphybased cladogram ( Sereno & Dong, 1992: fig. 14) that placed Huayangosaurus as the most basal known stegosaur. Several comments concerned unpublished comparative observations on the anatomy of Scelidosaurus and were used to justify exclusion of Scelidosaurus from the Eurypoda and reinforce Sereno’s (1986) topology ( Fig. 38 ).

4. Lee (1996) offered the first matrix-based analysis of ankylosaur relationships, but used a limited range of taxa and concentrated only upon cranial character-states. Similarly, Kirkland (1998: 273) analysed cranial and postcranial characters and doubled the number of taxa considered, but scored Scelidosaurus as the default outgroup in his matrix. Carpenter et al. (1998) undertook another analysis, but introduced suprageneric taxa ( Nodosauridae and Ankylosauridae ) that were determined prior to the cladistic analysis. The value of adopting an approach that makes a priori decisions about monophyletic groupings (suprageneric taxa) was criticized methodologically by Wilkinson et al. (1998) who determined that the results were phylogenetically uninformative. While strict application of logic renders their determination correct, it is the case that many of the more recent analyses of thyreophorans (and other, sometimes broader-based, taxonomic analyses) commonly make pragmatic a priori decisions about suprageneric taxa in order reduce data-processing time and ‘improve’ resolution. A ‘compartmentalized’ suprageneric taxon approach was nevertheless also employed by Carpenter (2001) in an attempt to manage the high levels of homoplasy that have been reported to occur among ankylosaurs (Penkalski, pers. comm. March 2020).

5. Sereno (1999) provided a more comprehensive review of dinosaur relationships by offering lists of characters (and their coding protocols) for each of the major dinosaur clades and their subordinate taxa. However, there were continuing methodological problems with this analysis through his use of suprageneric taxa (see Wilkinson et al., 1998): only a limited number of ornithischian taxa were included and the monophyly of particular named clades was assumed prior to the analysis. The thyreophoran dataset included 118 characters that were used to code 17 thyreophoran taxa. The clade Thyreophoroidea ( Scelidosaurus + Eurypoda) was supported by the five character-states listed above, but these had been modified in response to the discovery of a new, apparently more basal, taxon Emausaurus ( Haubold, 1990) .

6. Subsequent analyses by Vickaryous et al. (2001) and Hill et al. (2003) were more narrowly focused upon cranial characters alone and, as can be seen from the scores in the matrix produced by Hill et al. (2003: 28), it is not at all surprising that Scelidosaurus defaults as an outgroup to Ankylosauria (along with Emausaurus and the basal stegosaur Huayangosaurus ).

7. Vickaryous et al. (2004) provided what became the default phylogenetic analysis for ingroup ankylosaurs because it included cranial and postcranial characters, used a wide range of taxa and made no a priori assumptions about ingroup relationships. It was used for a number of years, with subtle modifications to character-states and the addition of new taxa. Unfortunately, this analysis used Lesothosaurus and Huayangosaurus as outgroups, and did not include basal thyreophorans such as Scelidosaurus and Emausaurus .

8. Butler et al. (2008) provided a re-assessment and more exhaustive set of analyses of the systematics of the clade Ornithischia . These were undertaken in the light of many new discoveries and the inconsistencies in the non-numerical systematic analyses that emerged from this new material ( Norman, 1984b; Sereno, 1984, 1986; Cooper, 1985; Maryańska & Osmólska, 1985). Butler et al. (2008) again used a number of suprageneric taxa (notably, in this instance, Ankylosauria ) and generated a larger ornithischian dataset than that presented by Sereno (1999). Regarding the placement of Scelidosaurus , their results were consistent with those published by Sereno (1986, 1999) because they used the same characters, apart from one additional character:

(a) The presence of cortical remodelling on cranial bones (after Carpenter (2001).

However, it was noted that Scelidosaurus had not been described (by 2008), with the implication that characters and character-state codings may differ once its anatomy is known in greater detail. It was also admitted that the analysis involved the coding of supraspecific operational taxonomic units, e.g. Stegosauria and Ankylosauria , and that a fuller consideration of the phylogenetic position of Scelidosaurus would (ideally) require consideration of a range of individual eurypodan taxa. However, this was considered to be beyond the scope of their analysis.

Overall, the analysis cast doubt upon the interpretation of Scelidosaurus as a stem ankylosaur ( Norman, 1984b; Carpenter, 2001). Constraining the resolved tree so that Scelidosaurus is positioned as the sister-taxon to Ankylosauria increased tree length by eight steps (among trees of 485 steps), but they did note that this was not a significantly worse explanation of the data (Templeton Test, P = 0.04–0.13).

9. Several more taxonomically restricted analyses of thyreophoran ornithischians have been undertaken since the work of Butler et al. (2008). The most relevant among these, because they incorporate Scelidosaurus , are Thompson et al. (2012), Arbour & Currie (2016 – with a supplementary by Arbour & Evans, 2017) and Wiersma & Irmis (2018) for Ankylosauria ; and Maidment et al. (2008), Mateus et al. (2009), Maidment (2010) and Raven & Maidment (2017) for Stegosauria . All of these studies provide taxon lists, as well as detailed character descriptions and coding.

The output of these latter sets of analysis differ markedly. For ankylosaurs, large numbers of equally most parsimonious trees (MPTs) were generated from data tables in which the codes assigned to characters were unweighted and unordered [4248 MPTs (52 taxa and 170 characters) – Thompson et al., 2012; 3030 MPTs (44 taxa and 177 characters) – Arbour & Currie, 2016]; and finally 21 MPTs [35 taxa (31 thyreophorans) and 293 characters – none of which were weighted, but 48 were ordered] – Wiersma & Irmis (2018). For stegosaurs, the datasets were smaller: Maidment et al. (2008) used just 18 taxa (11 of which were ingroup stegosaurs) and 85 characters, whereas the most recent analysis ( Raven & Maidment, 2017) used 23 taxa (13 of which were ingroup stegosaurs) and 114 characters. In contrast to the ankylosaur analyses, some characters were assessed (a priori) and selectively weighted or ordered. This procedure generated respectively five, 41 and finally a single MPT, seemingly considerably better resolved. Thorough though the analytical processing of all these studies has been, they are, somewhat paradoxically, limited with respect to their consideration of basal taxa (including Scelidosaurus ). As a consequence, in all instances but one ( Wiersma & Irmis, 2018), Scelidosaurus occupies an entirely consistent position as the sister-taxon to Eurypoda.

In each analysis, character lists are notable for their choice of codable anatomical characters intended to differentiate between a range of derived, but anatomically conservative and, in many instances, fragmentary/ incomplete ingroup taxa (note, for example, the discussion in Thompson et al. 2012: 308 et seq. regarding the status of the clade Polacanthinae/ Polacanthidae ). The cumulative effect of scoring large numbers of derived characterstates is that it introduces substantial statistical bias within the dataset that results in the outgroup OTUs and/or ‘basal’ taxa being scored ‘absent’/‘0’ for large numbers of characters (e.g. Butler et al., 2008; Thompson et al., 2012, et seq.) despite the fact that these characters enable resolution between more derived taxa within the overall analysis; these particular issues are discussed in greater detail by Brazeau (2011). Interestingly, Wiersma & Irmis (2018) compared the results of their analysis with those achieved by Arbour & Currie (2016 – and its reworked and heavily pruned iteration: Arbour & Evans, 2017). They noted a profound lack of resolution among most nodosaurid and ankylosaurid taxa in the strict consensus tree, and that a degree of resolution was only achieved by calculating a 50% majority rule consensus (which should not be used to explore phylogenetic relationships – Sumrall et al., 2001); and a maximum agreement subtree, which, although it identifies consistent phylogenetic structure common among the MPTs, also removes a large number of taxa and is inherently unstable ( Wiersma & Irmis, 2018). Weirsma & Irmis’ trees also lack resolution, although this does not appear to be as severe as that evident in Arbour & Currie’s data. Both studies highlight weak levels of tree support, which point toward fundamental characterrelated problems (high levels of homoplasy and missing data for many taxa) in ankylosaur systematics.

Scelidosaurus , Emausaurus and Scutellosaurus cluster at the base of most trees because they can be scored for only a restricted number of anatomical characters, many of which were identified originally by either Sereno (1986, 1999) or Butler, et al. (2008). Other factors that have influenced the phylogenetic reconstructions based on these analyses (notably the placement of Scelidosaurus but also including Scutellosaurus and Emausaurus ) are:

• The comparatively high levels of missing data for all three taxa.

• Anatomical characters were incorrectly identified and/or scored because so little detailed anatomy of these taxa was known.

In the case of Scelidosaurus , despite the skeleton having been cleared of matrix, authors have until now been obliged to rely on the monographs of Owen (1861, 1863), a small amount of anatomy illustrated by Charig (1972) or, on rare occasions, brief examination of material in the collections of the Natural History Museum (e.g. Carpenter et al., 2013; Arbour & Currie, 2016). An illustrative example of this particular problem can be seen in the matrix created by Arbour & Currie (2016) in which ~50 out of a total of 177 characters were scored incorrectly for Scelidosaurus ; this observation has no bearing on the competence of these authors but simply reflects how little was then known of this important taxon.

NEW ANALYSIS

For the present analysis, the author surveyed, assessed and sampled previously published character lists and their codings, then constructed a matrix of 15 taxa and 115 characters (Supporting Information, Appendix S2). The approach used was to identify characters that could be scored for currently known early (or apparently anatomically basal) taxa, as well as some exemplar well-preserved and well-described stegosaurs and ankylosaurs (see Supporting Information, Appendix S2). This resulted in the production of a near equal split between cranial and postcranial characters (cranial 55:60 postcranial). A number of characters were added, whereas others were redefined, corrected and re-coded. This process of winnowing avoided the incorporation of many highly specific characters that have no relevance to a consideration of basal taxon systematics ( Brazeau, 2011).

ANALYTICAL PROTOCOLS

Figures

Figure 38. Early (non-numerical) cladistics-based attempt to establish a topology for armoured (thyreophoran) dinosaurs within the clade Ornithischia. This topology is derived from Sereno (1986: fig. 3).