M. Osmond* and M. Wright^#

^*Advanced Infomancy Focus Group, Department of Alchemical Engines, Mercadia University Scholars, Mercadia; aethermail: infomancy@mus.org.uk and ^#Hall of Arcane Herpetology, Riptide Memorial Foundation, Nivix Collegium, Ravnica; aethermail: mjl.wright@gmail.com. Both authors contributed equally to this work and any correspondence should be addressed to both.

Running Title: Dragons in all shapes and sizes.

This paper concerns our initial attempt to categorise all known members of Draconis into a simple taxonometric dendrogram. This process involves first describing the physical characteristics of each member according to a fixed list of attributes or ‘traits’ (e.g. number of wings or feet, leathery or scaly skin, known mana affinities, etc.). This information was then collated and machine analysed according to Pearson’s Metric of Similarity. The resulting model thus separates the family into groups according to resemblance, with early splittings in the tree indicating significant physical and arcane differences between major types (e.g. between artificial and ethereal varieties). The authors hope that this entirely trait-orientated (and hence plane-agnostic) approach may shed some light on the actual evolutionary development of these creatures.

Long the subject of fascinated study by almost all known civilised races, the members of family Draconis have inspired fear, awe and no few attempts at imitation (1, 2) across planes and eons alike. The huge variation in their morphology, behaviour and preferred habitats, as well as the inherent and obvious difficulties in studying these creatures, has seriously impeded progress at description and categorisation let alone the development of a rigorous taxonomy.

Figures 1 & 2. Representative examples of (left) D. nobilis and
(right) D. vulgaris.

While a number of possible Draconis genealogies have been proposed (3), all are heavily reliant on oral historical evidence which is by nature suspect. In the absence of a description of ancestral line from a primary source – which seems unlikely to ever occur – genealogies of this type will remain embroiled in endless debate. Even at the simplest level, it remains unclear whether the classical division between the species-groups D. vulgaris or ‘drake’ and D. nobilis or ‘true dragon’ has any biological basis, or is merely the result of convenient taxonomic pigeonholing (4).

Complicating the issue further, the distinction between draconic races and distinct individuals becomes confused when working from (often fragmented) historical records. This is particularly true in reference to the so-called ‘Elder Dragons’, beings of such age and power that they may reasonably be considered the ultimate ancestor for entire draconic populations (5). In an attempt to avoid confusion, the authors have decided to treat all reported Draconis morphologies equally and have analysed them with respect to their physical appearance and innate arcane aspect, with minimal inherent preconceptions and with every attempt made to work from reliable and established sources (6).

The classical system of draconic classification has suffered a number of significant upheavals in the last few years. In particular, reports of non-corporeal dragons (originally misidentified as phantoms but now confirmed as spirits by the excellent work of the Minamo School researchers) have forced the creation of the grouping D. arcani to contain them (7). This joined the existing groups: D. vulgaris, normally described as small or medium-sized, two-legged, and blue in aspect (Fig. 2); D. nobilis, large, red and four-legged ; and the hotly disputed D. wyvar. Some authorities also lobby for the official addition of D. facticius and D. inferi to contain artificial and undead varieties, although these are arguably constructs rather than actual organisms (8).

This approach may be considered a classification by variation from the ideal or ‘perfect dragon’, which is normally described as four-legged, winged, scaled, fire-breathing, and with a red aspect – the standard example being the great Fire Wyrms of Shiv (9)(Fig. 1). This approach has been recently criticised as “small minded and pretty damn stupid” by a notable expert (10) who called for a more logically structured and extensible system, better able to describe new races as they are reported without the need for the current myriad of special cases.

Towards this aim the authors have pursued and collated descriptions of all reported dragon variations (both extant and historical). Wherever possible the best visual description of each phenotype was used, with a preference for primary sources (11) or reports corroborated by multiple sources (12). These examples were then classified using a series of physical characteristics, e.g. leg and wing count, scaly v.s. smooth skin or the presence of frills, as well as a number of more general characteristics, e.g. mana affinity or ethereal nature. Unfortunately the number of easily identifiable traits is limited. While the authors initially supported attempts to include more intimate physiological characteristics (for example, blood type and tooth shape), a number of unfortunate incidents during the data collection process lead to our decision to limit characterisation to traits visually identifiable at a distance.

Experimental Details

Data collection – the traits chosen for characterisation were: mana affinity; the number of wings, heads, and legs; skin texture (scaly, leathery, metallic); the presence of neck, cheek, and tail frills; the presence of a beard or horns; colour (white/yellow, green, blue, purple, red, orange, brown, or combinations of these) and more unusual natures (ethereal, artificial, or undead). Data from all traditional branches of Draconis were combined and analysed together.

Infomantic analysis – the trait data was examined using a Kjeldoran Daemonic Enviroment (v1.8) running in a PyrEX III calculating engine (Yawgmoth Corporation, Rath). After initial parameterisation, the process was entirely automated with no further intervention by the authors. It should be noted that the resulting groupings are therefore entirely a result of the trait characterisation of each dragon; as is standard in such analyses the daemons participated under a level 3 impartiality geas.

Discussion

We will start by briefly summarising the results of the hierarchical clustering, the dendrogram for which may be seen in Figure 3 (readers with access to the aethernet may wish to examine the interactive illustrated version at http://www.mus.org.uk/clusterer/original/).

Figure 3. Proposed taxonomic dendrogram with special
features marked.

The analysis algorithm iteratively examines each group of dragons and splits them into two sub-groups or clades, such that each clade contains members that are most similar to each other and most different from the other sub-group. This process can reveal which features are useful for separating Draconis into these clades, and the hierarchical tree gives some indication of their relative significance.

We will now briefly discuss the results of the hierarchical clustering and our conclusions from them.

Artifact and Ethereal – The very first division in the dendrogram (a) is perhaps the most indicative and satisfying. It is a highly unbalanced splitting, with the majority of the ethereal and all the known artificial dragons being separated from their flesh and blood cousins. Division (b) then sees these two variations further separated into clades equivalent to D. arcani and D. facticius which appears to be a strong argument towards formal acceptance of the latter.

A notable (and probably erroneous) oddity is the position of the distinctly non-ethereal Jamuraaian Canopy Dragon (c)(Fig. 4). Current thinking explains the creature’s unique morphology (wingless with 6 legs), as an adaption of an ancestral 4-leg, 2-wing form for better mobility within the dense jungle canopies of Mwonvuli (13). By contrast, the ki dragons with which it is grouped are entirely magical in nature and appear to be a form of Kami. There is however a clear physical resemblance between the Canopy and the rest of the clade (all of whom are also wingless but able to fly) and so the daemonic confusion is understandable.

Figures 4 & 5. Examples of oddities and misappellations; (left) the Canopy Dragon and (right) the Pendrell ‘Drake’.

The Undead – Previous classifications have normally treated resurrected Draconis as a special case (3, 8). Indeed, the recent call for the inclusion of D. inferi assumes that they form a distinct clade. It is therefore surprising to report that our analysis did not group Bladewing the Risen (d), the Vampiric Dragon (e) nor the Tattered Drake (f) together. Instead these are placed in clusters mixed with non-undead dragons.  As the undead physiology does not differ greatly from that of living dragons (albeit somewhat degraded) this entirely logical. It is also interesting to note that Rorix (both alive and undead) has a unique morphology of four wings, and so in both forms creates another early branch in the hierarchy.

Swamp Draconis – Amongst the living Draconis, the next identifiable cluster contains individuals sharing a tendency to black or dark purple colouration (g). This trait is not unique to the clade however, so this cannot be the main influential feature.

The five most similar members are all pure black (an uncommon colour-form). In addition, the Double Header and the rare  ‘Portal’ form of the Alabaster Dragon both show partial melanistic colouration as well as unusual morphology: the Header has two heads and two tails, and the Alabaster has six legs (3). Since all members of this clade are known to favour fens or swampy environments it could be speculated that these species are the result of the mutagenic influences common to these habitats.

Drakes and Dragons – The most significant split (h) in the main branch is based upon the number of legs and maps directly onto the classical split between D. nobilis and D. vulgaris. In this aspect the classical model seems to hold strong, with the ‘dragons’ cluster containing only members with 2 wings and 4 legs, whereas those in the ‘drakes’ cluster have only 2 legs. Indeed the clearness of this split has allowed the authors to easily identify a number of misappellations in the historical record. For example, the ‘Portal’ form of the Wind Drake should be correctly classified as a small dragon. It seems likely that the same also applies to related Draconis such as the Pendrell (Fig. 5), Wormfang and Canyon ‘Drakes’.

The authors argue that the established tendency to name small Draconis as ‘drakes’ irrespective of morphology is both confusing and disrespectful (similarly to the now well-discredited conception that ‘All dragons are red and all drakes are blue’ (14). Readers on the aethernet are encouraged to access our interactive dendrogram in order to better comprehend the true body shapes without having to rely on the traditional, sadly inaccurate, names. It should be noted that a few exceptions still exist, for example the Iridescent Drake (i) is clearly legless and the ‘Mirage’ Wind Drake (j) is placed in the ‘drakes’ cluster despite having four legs.

Further Subclassification – For both dragon and drake clusters further subdivision into clades appears to be influenced predominantly by body-colour, and subsequently by skin type. The drakes split off into a small group of orange drakes (Flailing Drake, Desert Drake, Volcanic Dragon – a clear misnomer!, and Flowstone Wyvern – (k)) and a large group of primarily, but not exclusively, blue drakes (l). The dragons may then segregate into four main races: the Clouds, the Aboriginals, the Primevals and the Shivan.

The Clouds (m) are a combination of two smaller clades with a shared affinity for white or blue, commonly associated with the sea or meteorological features. Their morphology is highly variable and representative traits include vapourous bodies, silvered skin and a general physical robustness. Arcades Sabboth, Chromium, and Treva, the Renewer are probably the most notable members of this race. After the Clouds is a small, well-separated clade that we have designated the Aboriginals (n). Members of this race are normally orange-red or green, and have an extremely classical body shape with minimal additional features such as frills. They appear to have a preference for ‘fringe’ habitats and are commonly found living on the edges of islands, mountainous regions and deserts. The authors suspect that the Aboriginals represent some of the oldest and most primitive (in evolutionary terms) of the dragon sub-species.

Figures 6 & 7. The Primevals are some of the oldest and most powerful members of family Draconis; (left) Darigaaz, the Igniter and (right) an unidentified Ethereal.

The last two races include many of the most notorious and respected draconian legends and subspecies. The Shivan (o) are the largest of the draconic races (both in physical size and number) and a full discussion of them is beyond the scope of this work. However it should be noted that this clade is centred on the Shivan Dragon (considered by some to be the epitome of classic draconic form (9)) and contains within it such illustrious members as the Kilnmouths, the Rathi, the Tyrants and the Hellkites. The Primevals (p) share many Shivan characteristics (in particular their size and power) but show far greater variety of individual pigmentation and ornamentation. Indeed these traits (especially in the case of magically Adept members) appear to be largely a matter of personal preference. The Primevals are epitomised by the ‘Elder Dragons’, the bright and terrible beings from Dominaria’s ancient past. The Shivan Dragon and the Elders are extremes and most members share characteristic properties from both races. A excellent example is the small grouping (q) containing the Imperial Hellkites and Niv-Mizzet, whose cladistic proximity is supported by their clear family resemblance and suggests a possible solution to the Cromat enigma.

Further Considerations – This model, whilst it should by no means be considered an authoritative illustration of dragon ancestry, may add further support to existing namings/identifications, or conversely suggest that further research or reclassification may be productive. To demonstrate, the authors would like to highlight some similarities and differences of particular note.

The Draconis trait descriptions were collated from sources of great spatial, temporal and planar variety. It is therefore gratifying to see that both the ‘Alpha’ (described circa 200 AR) and the ‘Seventh’ (described 4301 AR) forms of the Shivan Dragon are placed adjacent to each other. Similarly, the Spiketail Drake and Hatchling are clearly related, as are the Furnace Dragon and Whelp, while Dromar, the Banisher appears to be a giant Tempest Drake. Continuing in this theme, the Drake Hatchling may well be of the Tolarian sub-species and the Izzet Familiar appears to be a variety of Timid Drake (which may explain the relative ease of domestication). The Benalish Tower Drake and Ravnican Snapping Drake may actually be members of the same sub-species – as well as similar morphology, both have a well-documented preference for urban habitats. Finally, the authors would like to highlight the significant variances in the many Azure, Snapping and Fighting Drake sub-species, which probably arose from a historic tendency to name domesticated drakes by their function or nature rather than their genealogy.

Conclusions and Future Work

We have described our approach to systematically generating a taxonometric dendrogram of the known draconic forms and provided our interpretation of the clades so produced. This hierarchy has already been helpful both in identifying likely misclassifications and confirming existing relationships. We hope it will also be of value to inform the debate on the composition and naming of draconic families. We strongly corroborate the classic D. nobilis / D. vulgaris split, and believe this work provides good support for the currently controversial D. arcanis and D. factitius, but not for D. inferi. The authors are of the opinion that the “perfect dragon” model still has value, in that Shivans form the largest (and most awe-inspiring) cluster, but we hasten to note that trait variance in this clade is very high and therefore alternatives to the Shivan may prove to be more appropriate. We also note that this model demonstrates a potential approach for the morphological back-extrapolation to the primordial, or Ur-dragon, but will leave this endeavour to others.

This study should however not be considered a completed work. The study of Draconis is by its very nature ongoing and descriptions of new sub-species (both dragon and drake) are published regularly. Certainly no authority can claim to have a complete understanding of draconic range and variety, hence any taxonometric system must in turn be flexible enough to take this into account. The authors feel that the model described herein forms an excellent foundation for such a system. It is our intention to expand this existing model, not only with respect to new Draconis members but also to the number and variety of traits analysed. By adopting a more detailed trait description and by increasing the number of samples to include draconians mentioned peripherally in other records, we hope to significantly improve the quality and stability of the cladistic tree generated.

Acknowledgements

This study would not have been possible without the tireless efforts of many individuals. In particular, the authors would like to express their gratitude to the Avian Airforce Dragonstalker Corps., whose skills and dedication have collected some of the most detailed observations of draconic physiology and behaviour yet described. We would also like to thank Royal Empress Llawan for her gracious patronage and encouragement of the Riptide Project, the esteemed Parum Niv-Mizzet for financial support and many valuable (albeit heated) discussions, and Uktabi Infomancy Systems (Dominaria) for their generous donations of equipment and technical support.

Finally, and most importantly, we must thank the Wizards of the Coast (www.wizards.com) for their maintenance of the longest-standing trans-planar library, whose curators have gathered and preserved the original source material which made this analysis possible.