Monday, February 8, 2010

How Not To Go About Classifying Marine Vertebrates... And Cadborosaurus

Again with the Cladistics!
In a couple prior posts, here and here, I critically examined a well-known cryptozoological categorization and found that one proposed cryptid 'type' cannot be supported from a qualitative or quantitative standpoint. Heuvelmans' system of classification and its progeny appear to use methodology inspired by evolutionary taxonomy, an 'art form' approach which relies on perceived similarities and differences between taxa (Cracraft 2006). As evolutionary taxonomy has been superseded by cladistics in the last few decades, why haven't those interested in cryptozoology attempted a modern-style parsimony analysis of anecdotal information? There is a preconception that cladistics can only be used to determine the relationships of organisms, which was the original intent, but its principles can be applied to any hierarchal (i.e. groupable) data set (Young 1995). Another preconception is that data cannot be gleaned from anecdotal information, but it is possible so long as the conclusions drawn are limited (Paxton 2009). Cladistic analyses create a large number of trees, making it possible to determine bootstrap values, i.e. how often a certain clade is recovered; I suggest clades above 50% (majority) should be viewed as potentially informative and those with values of 95% (= p value of 0.05) and up can be considered data. Of course, even clades recovered in 100% of trees should be qualitatively reviewed in order to determine if the relation is just a quirk of sampling and/or characters used. 

In my Many-Finned analysis, I found the majority of anecdotal reports vague to the point of uselessness, strongly suggesting that any sort of grouping for this 'type'  is a priori and self-deluding. I understand that cryptozoologists would be wary of a method which recovers many reports as 'noise' and cannot be used to obtain 'proof', but cladistics is a potentially useful method to objectively and scientifically approach unusual information. Cladistics is regularly utilized to determine the molecular phylogeny of purported cryptid material (see Crother et al. 2009), and I see no reason not to expand its usage to other information present in cryptozoology - aside from its somewhat time consuming nature.

Caddy of Cascadia
LeBlond and Bousfield are investigators of a marine cryptid from the Pacific Northwest with the unfortunate moniker of "Caddy" or Cadborosaurus. While their work on collecting reports is commendable, as some are quite compelling, their scientific analysis of the purported species has been heavily criticized and controversial to the degree of causing two editors to resign from the amphipod journal (!) in which a species description was published.

For the sake of argument, we are going to have to assume that the authors had good reason for thinking that 'Caddy' displays the traits they listed, and has a strong enough chance of being a valid species that an attempted phylogenetic analysis would not be a waste of time. These are probably unwarranted assumptions and I'll be upfront that my primary reason for writing this post is to establish a cautionary tale for those interested in unusual anecdotes.

The authors used a form of analysis which was previously utilized in a study of amphipod systematics by LeBlond and another author (which I cannot access), and was justified in the context of classifying cryptids because "[t]he data are far too sparse and soft to use the more mathematical techniques of biological classification" (LeBlond and Bousfield 1995). Needless to say, I am highly dubious about the reasoning behind choosing a special method because the data cannot stand up to more conventional analyses! I have not seen any subsequent uses of the 'P/A Index' in peer-reviewed literature, which in itself is probably quite telling about its applicability.

After LeBlond and Bousfield (1995). Click to enlarge

8 Characters for 8 Clades!
The 'P/A Index' uses 8 broad traits to classify 'Cadborosaurus' amongst marine vertebrates and whatever 'Primitive Mammals' are. The incremental index values (bony fish 3+, amphibians 4+, reptiles 5, 8+, mammals 7, 11, 13) are suggestive of the 'Great Chain of Being', a bizarre Western concept which still pervades popular conceptions of evolution. However, a 'Plesiomorphic/Apomorphic Index' is theoretically possible if the characters used are highly unlikely to have evolved multiple times - non-craniate chordates and craniates can be distinguished on the basis of neural crest cells, non-sarcopterygian vertebrates and sacropterygians on the basis of lobe fins - and so forth - which can create a 'chain' of increasingly derived clades. I cannot see how this index method confers any advantages over cladistics. The authors only used one trait which is highly unlikely to have evolved multiple times, placental reproduction, but, as nobody has reported the presence/absence of an afterbirth, its coding is due to an a priori assumption on the part of the authors.

Before I cover the next enormous problem in LeBlond and Bousfield's index, what happens if take the numbers given in the index at face value and put it through PARS? This tests the assumption that the value given in the 'P/A Index' is at all predictive of phylogeny:


The next issue is that the authors used enormous clades consisting of thousands of species, with their variance expressed by '+' and '-'. What happens if we assume that each has a value of 1?

Cadborosaurus: 9-10?
Bony Fishes: 3-5
Amphibians: 3-7
Modern Reptiles: 5
Plesiosaurs: 8-11?
Primitive Mammals: 7-8
Pinnipeds, Advanced Mammals: 9-12
Cetacea: 13

Great Chain of Being no longer! I'm highly dubious about a system which gives similar scores for pinnipeds and plesiosaurs and cannot reliably distinguish fish, amphibians and reptiles. So can the P/A index be salvaged, or is it fundamentally flawed?

The Index Recoded
LeBlond and Bousfield had some very odd choices for coding character states, in fact, I would have done over half of them differently! Some of the problems were very troublesome, most notably that the small head/long neck condition was considered plesiomorphic, which is obviously not the case for a group which includes bony fish! Rather than go through all of the problems and offer a possible explanation, I'll define/refine the possible character states.

Total Length: Was 'Body Size'. Animal size should be expressed in terms of mass, but clearly this is not possible as we are dealing with a cryptid. So let's go by orders of magnitude: State 0 = 0 to 1 m, State 1 = 1 to 10 m, State 2 = 10+ m. Note that State 0 is what I consider to be plesiomorphic, as this is the state for Branchiostoma virginiae (a lancelet) one of the invertebrate chordate outgroups I shall be using.

Fineness: Was 'Body Shape'. 'Elongate' body shapes are considered to be over 7 (see this discussion), and I shall use 14 as the cutoff point for what I shall consider 'serpentine'. Both of my outgroups (a lancelet and a hagfish) are elongate, so I shall assume this was the plesiomorphic condition for vertebrates. State 0 = elongate (fineness 7 to 14), State 1 = compact (fineness under 7), State 2 = serpentine (fineness over 14). One potential issue is that animals which have partial serpentine-like morphology can code as elongate, e.g. Elasmosaurus has a fineness ratio of only 10.2.

Neck Length: Was 'Head and Neck'. State 0 = no neck, State 1 = neck present and of similar/lesser length than head. I shall also accept pseudo-necks present in some fishes, which I shall define as portions of the body which are noticeably more slender than the head (e.g. narrowneck eel, sea horse). State 2 = neck far longer in length than head, comparable to swan or plesiosaur. This is a bit more nebulous than I would like, but should serve us well for the taxa considered.

Paired Appendages: Formerly 'Number of Legs'. The name was changed in order to account for the pectoral and pelvic appendages of fishes (which the authors considered legs...). State 0 = appendages absent, State 1 = 1 pair, State 2 = 2 pair. The lack of appendages is a plesiomorphy as evidenced by the lancelet and hagfish. Unlike LeBlond and Bousfield, I still count appendages if they are reduced or have a novel orientation.

Caudal Appendage: Formerly 'Tail Style'. The name was changed in order to account for appendages located near the tail (e.g. hind flippers of pinnipeds) that could look as if they were bilobate. State 0 = vertical caudal appendage, State 1 = no caudal appendage, State 2 = bilobate or bilobate-like caudal appendage.

Thermoregulation: Formerly 'Thermal Physiology'. State 0 = Ectothermic regulation, State 1 = Homeothermy via Rete mirabile, 'Gigantothermy', et cetera. Bradymetabolic animals. State 2 = Endothermy (heat via metabolic rate), Homeothermy (stable internal temperature), Tachymetabolism (high resting metabolism).

Reproduction Location: Not where the sex occurs, but where young are spawned/born. State 0 = Aquatic, State 1 = Terrestrial (as depicted by LeBlond and Bousfield, there's no way 'Caddy' could haul out on land to reproduce).

Reproduction Type: State 0 = Oviparity, State 1 = Viviparity and Ovoviviparity (lumped to eliminate a priori reasoning).

Now with the characters more sensibly defined, we need to abandon the huge clade categorizations and use individual taxa. I'm using 30 taxa which should provide a reasonable sample of the diversity of marine and semi-aquatic vertebrates. I specifically went for large and morphologically unusual species:

Branchiostoma virginiae (Amphioxiformes)

Eptatretus goliath (Myxini)

Huso huso (Acipenseriformes)
Anguilla anguilla (Anguilliformes)
Derichthys serpentinus (Anguilliformes) - been meaning to blog about this one for a while
Poecilia reticulata (Cyprinodontiformes)
Monopterus indicus (Synbranchiformes)
Regalecus glesne (Lampriformes) - I am not certain if they can actually reach 10 meters.
Benthodesmus tenuis (Trichiuridae)
Thunnus thynnus (Scombridae) - yes, it has regional endothermy.

Bufo marinus (Anura)
Andrias davidianus (Caudata)
Amphiuma tridactylum (Caudata)
Chthonerpeton viviparum (Gymnophiona)

Dermochelys coriacea (Chelonia) - an endothermic species which even has mammal-style blubber.
Chitra chitra (Chelonia) - I have no idea how this neck can contract (and it does...)
Varanus komodoensis (Squamata)
Eunectes murinus (Squamata)
Hydrophis spiralis (Squamata)
Crocodylus porosus (Archosauria)
Cygnus olor (Archosauria)
Mauisaurus haasti (Plesiosauria)

Ornithorhynchus anatinus (Monotremata)
Ondatra zibethicus (Rodentia)
Pteronura brasiliensis (Carnivora)
Eumetopias jubatus (Carnivora)
Hydrurga leptonyx (Carnivora)
Phocoena phocoena (Cetacea)
Lissodelphis borealis (Cetacea)
Balaenoptera physalus (Cetacea)

'Caddy' (Unknown)

And their character states:
Click to enlarge

Note that giant salamanders, sturgeons, eels, and eel-like fishes have the same score (4); as do caecilians, oarfish, tuna, and crocodiles (6); softshell turtles and platypi (8); muskrats and anacondas (9); and so forth. It's clear that this index can only be used for the broadest of classifications (fish vs. mammals) and that 'Caddy's' score is essentially meaningless. Of course, there's still one thing I'd like to do:

...and that's why you don't use 8 homoplastic characters to determine the phylogeny of vertebrates!

After subjecting the data to the regular regiment, a phylogenetic tree emerges that is about as traumatizing as watching J. Walter Weatherman getting his arm unexpectedly and repeatedly amputated at several points throughout my childhood. My favorite is the marine toad/guppy clade.

Piecing together dissociated knowledge to open up terrifying vistas of reality

  1. As depicted by LeBlond and Bousfield (1995), 'Cadborosaurus willsi' resembles a secondarily aquatic vertebrate - but that's about the limit of what can be said.
    1. Their depiction appears to be heavily based on the Naden Harbour carcass.
    2. There is a distinct possibility that said carcass is that of a basking shark
      1. Has anybody ever commented on how featureless the 'skull' is and all those weird projections coming out of the neck? 
  2. LeBlond and Bousfield (1995) classified 'Caddy' amongst the marine vertebrates using 8 characters.
    1. It might be possible to get a crude topology of vertebrates using 8 characters (e.g. presence of lobe-fins, amnion, dentary-squamosal jaw joint, avian-style wings, et cetera) which have almost certainly evolved once.
      1. None of these traits are observable/inferable from 'Caddy' reports. 
    2. Almost all of the traits the authors used have evolved numerous times, dozens for some.
    3. The characters they used were vaguely defined.
    4. The authors compared 'Caddy' with different vertebrate 'classes'.
      1. Those groups are composed of thousands of species.
      2. Their character states cannot be accurately represented with a number and a character.
    5. LeBlond and Bousfield (1995) had some very questionable coding:
      1. According to the P/A Index:
        1. All 'Modern Reptiles' are serpentine and have long necks
        2. 'Modern Reptiles' and 'Bony fish' are always 'cold blooded'.
        3. Plesiosaurs occasionally have less than 4 limbs and horizontal tail flukes.
        4. Pinnipeds have a single pair of limbs and sometimes more
        5. Cetaceans, 'Modern Reptiles', and Amphibians have the same number of limbs.
        6. Et cetera.
  3. LeBlond and Bousfield (1995) count up the number of apomorphic characters as if that means something.
  4. 'Cadborosaurus' placed ambiguously (9?)
    1. "Overall, Caddy ranks most closely with marine saurischian [sic] (plesiosaurs) or thalattosuchian (marine crocodilians) reptiles and marine mammals, which is perhaps not too surprising. To be more definite, one has to weigh the relative importance of the various characteristics." (LeBlond and Bousfield 1995, p.82).
      1. Saurischia = Theropoda (including Aves) + Sauropoda
      2. Read: The traits that give the desired identification will be weighted.
    2. The authors make the following arguments for why 'Caddy' is a reptile:
      1. Reported hair may be "something else" 
        1. [Pseudo-hair? What about reported whiskers?]
      2. While "no living reptiles are capable of bending their spines in the vertical plane" - marine crocodiles had whale-like vertebrae and were capable of vertical flexure
        1. [Assuming the latter point is even true, the authors do not understand the difference between vertical flexure and vertical undulation. Not even all mammals locomote with vertical undulation.]
      3. Slenderness and implied considerable surface area in cold waters implies that 'Caddy' is 'poikilothermic'.
        1. [Yet they claim it is an extremely fast swimmer and can fold itself into a compact form comparable with tunas and orcas!]
      4. Reported precocial young were very small in comparison with adults like reptiles.
        1. [Extremely wrong. This shall be answered elsewhere.]
  5. On the plus side, the authors don't there there is enough evidence for multiple cryptids behind 'Caddy'.

Speculating on the phylogenetic relations of 'Cryptozoological Mega-Monsters' like 'Caddy' which have highly debatable traits (Read: chimerical, poorly supported, and questionably interpreted) is far beyond the applicability of the information available. I seriously question the statistical reality of marine cryptid 'types', as reading through 'Caddy' reports in LeBlond and Bousfield (1995) reveals them to be only slightly less vague and polymorphic than the Many-Finned. And this is likely the best-supported 'type' of marine cryptid! I would strongly suggest that sea serpent 'types' should be jettisoned unless thorough qualitative and quantitative analysis suggests otherwise.

This does not mean that we should stop investigating sightings and strandings. Some of the more detailed reports can be subjected to qualitative and cladistic analysis to determine if an identity can be established. However, if somebody like Bob Pitman sees an unidentified beaked whale (which he did) and only mentions a handful of characters - considering the unambiguous relation with an extant clade and the source of the report, we can seriously propose this as a cryptid. The only way to establish that the 'Cryptozoological Mega-Monster' is out there is to wait for an unambiguous carcass to show up, and I'm sure everyone familiar with cryptozoology is aware of how that has gone so far...


Cracraft, J. (2006). Why Classifications are Essential Tools for Comparative Avian Biology. Wings Without Borders: IV North American Ornithological Conference. 71. Available.

Crother, BI, et al. (2009). Giant Canadian Snakes and Forensic Phylogenetics. Contemporary Herpetology 2009(2), 1-4. Available

LeBlond, PH and Bousfield, EL. (1995). Cadborosaurus Survivor from the Deep. Horsdal and Schubart Publishers: Victoria, Canada.

Maddock L, et al. (1994). Mechanics and physiology of animal swimming. Marine Biological Association of the United Kingdom: United Kingdom. Partially Available.

Paxton, CGM. (2009). The plural of 'anecdote' can be 'data': statistical analysis of viewing distances in reports of unidentified large marine animals 1758–2000. Journal of Zoology 279(4), 381-387.

Perrin, WF, et al. (2008). Encyclopedia of marine mammals. Second Edition. Academic Press: New York, USA. Partially Available.

Young, GC. (1995). Application of cladistics to terrane history - parsimony analysis of qualitative geological data. Journal of Southeast Asian Earth Sciences 11(3), 167-176.


Anonymous said...

Opulently I assent to but I contemplate the brief should have more info then it has.

Cameron McCormick said...

Is this a Turing Test?

Mickey Mortimer said...

Interesting post. The Cadborosaurus paper is clearly terrible. You should order the first three characters though, changing the states to be in logical order - in cladistic analysis it doesn't matter if the outgroup always has state 0 (for the slenderness character).

Cameron McCormick said...


Those strange values are a holdover from trying to build a better P/A Index.