Loss of resolution in cladograms

A cladogram
is a graphic image (a diagram) typically generated from software, and typically based on a list of taxa and a list of characters, each one scored for each taxon. The cladogram is supposed to model actual evolutionary relationships among organisms. Ideally every cladogram will be fully resolved with every taxon sufficiently different from its sisters to merit its own node or branch. This is the basis for lumping and splitting. In practice loss of resolution occurs when sisters are not sufficiently different from one another. This can happen due to one or more of several problems:

  1. Three taxa are in reality too closely related. As an example, you may have input several specimens of the same genus and species, like “Tyrannosaurus rex,” without having characters in your characters list that could taxonomically separate them. Two identical taxa nest together, no problem. Three identical taxa produce loss of resolution. We saw something like this earlier in pterosaurs were two Rhamphorhynchus specimens had the same score and were deemed to be an adult and juvenile of the same species – but the third sister taxon (of the same putative species!) did not have the same score.
  2. Two sister taxa share no characters in common despite being closely related. This occurs most often when a skull-only taxon nests as a sister to a skull-less taxon, but it could also occur with other combinations of missing parts.
  3. One taxon is in a ‘by default’ nesting. It should not be in your taxon list because in reality it is not closely related to any other taxon in your taxon list. For instance, when one attempts to nest the pterosaurs “Pteranodon” or “Dimorphodon” with generic and specific members of the Archosauria or Archosauriformes, or when one attempts to nest “Homo sapiens” with generic and specific members of the Ichthyosauria. No telling what can happen in those instances, but anyone is able to try.
  4. Too few characters #1. If you have only 5 or 50 characters, there may not be a large enough list of traits to split your taxa apart. Statistically the list becomes large enough for 98% certainty at around 150 characters and becomes incrementally better with every character added thereafter. The large reptile tree uses 228 characters, many with more than two options, and has complete resolution (except for skull-less and skull-only taxa nesting as sisters (#2 above) with 546 taxa. Thus in practice there is no 3:1 character:taxon ratio as you may have learned about in theory.
  5. Too few characters #2. If your sister specimens are only represented by a few bones or parts of bones then two sister taxa may not resolve.
  6. Too few characters #3. Legless burrowing tetrapods appear to converge in their remaining traits so you better have a sufficient number of traits to lump and split the various clades. Otherwise the legless clades tend to be attracted to one another.
  7. Too few taxa. This goes back to #3 above because by throwing in one ‘by default’ taxon, like “Vancleavea” into an unrelated clade, like “Archosauriformes,” without also including the verified sisters of “Vancleavea,” like “Helveticosaurus” and “Askeptosaurus,” might produce loss of resolution because “Vancleavea” shares so few traits with any archosauriforms and the addition of the other thalattosaurs will clarify relationships. You may not have loss of resolution, but by adding taxa you’ll eliminate these ‘by default’ taxa.
  8. Using suprageneric taxa. If you can pick traits from one partial specimen AND another partial specimen to get enough traits to fill a suprageneric character list for a single taxon, then you’ve created a chimaera that can only lead to trouble. Even if your two taxa are incomplete, that’s better than creating a single cherry-picked chimaera taxon.
  9. Mistakes in scoring. Since humans are scoring all characters, mistakes can happen and they can affect nestings. Mistakes are often due to: 1) trying to maintain a paradigm or tradition; 2) too much leeway or opinion possible in a choice of possible scoring options; 3) inadvertent transpositions of data; 4) typos; 5) relying on the veracity of prior scorings, etc. Double and triple check your work and the work of others when you find loss of resolution. Errors are easy to make.
  10. Loss of resolution can occur at several levels: 1) in a simple heuristic search you need only one character score to lump and split sister taxa from your list of several dozen to several hundred traits; 2) in a bootstrap/jacknife search you need at least three character scores to lump and split taxa to raise your bootstrap score over 50%.

In my manuscripts,
when I report that my trees are fully resolved, that never seems to impress the referees (or they don’t believe it). Perhaps that is so because so many accepted manuscripts have loss of resolution at several nodes for many of the above reasons. That is not acceptable in most cases (exception: skull-only taxa will continue to occasionally nest with skull-less taxa).

We know better now.
We now have a large gamut “umbrella” study that continues to recover relationships within the Reptilia as it continues to increase in size. This large study provides a basis for smaller, more focused studies. When the old unverified traditions and paradigms have been replaced with verified models and relationships, then we’ll all have more confidence in recovered trees.

Advertisements

Tetraceratops news

A paper by Spindler (2014) sought the oldest therapsid…
and found no reason to nest the former best candidate, Tetraceratops (Fig. 1), with basal therapsids, confirming what was reported here in 2011.

New reconstruction of the limnoscelid Tetraceratops.

Figure 1. New reconstruction of the limnoscelid Tetraceratops.

From the abstract: “Since the successful clade of therapsids occurs rather suddenly in the fossil record of Guadalupian age*, the reconstruction of their origin is questionable and based on little data. Concerning the Artinskian taxon Tetraceratops insignis**, broadly accepted as the oldest and basal-most member, no close relation to therapsids could be found during the re-documentation. Instead, a fragmentarily preserved vertebral sequence from the Desmoinesian assemblage [Late Pennsylvanian, Westphalian D] of Florence, Nova Scotia, is considered to be a new candidate for the oldest therapsid. This pushes back their origin farther than required by phylogenetic results. Moreover, it supports the ghost lineage of unknown Carboniferous and Early Permian therapsids.”

* The large reptile tree nests the basalmost therapsid, Cutlieria, in the Early Permian.
** The large reptile tree nests Tetraceratops (Early Permian) with the limnoscelid, Tseajaia.

Spindler reports, “The problems when evaluating Tetraceratops are (1) its highly autapomorphic character combination, such as ornamentation, short facial region, and specialized dentition, and (2) the poor preservation of the single holotypic skull. The specimen has been re-studied carefully and is currently under re-evaluation. Anatomical identifications take into account a high degree of compaction, but although a simple mode of deformation. In contrast to previous workers, the therapsid synapomorphies could not be reproduced, resulting in a haptodont-grade classification*** independent from the same result by CONRAD & SIDOR (2001) and supported by LIU et al.(2009).”

*** Spindler did not consider a larger taxon list that included Tseajaia and limnoscelids.

Unfortunately
I cannot nest a vertebral sequence in the large reptile tree, so until I can (probably never), I accept Spindler’s observations and interpretations.

References
Spindler F 2014. Reviewing the question of the oldest therapsid. Paläontologie, Stratigraphie, Fazies (22) Freiberger Forschungshefte C 548: 1–7.

Ianthodon: a basal edaphosaur without tall neural spines

A new paper
by Spindler, Scott and Reisz (2015) brings us new data on the basal pelycosaur Ianthodon schultzei (Fig. 1; Garnet locality, Missourian Age; 305-306 mya, Middle Pennsylvanian, Late Carboniferous). The authors reported that Ianthodon represented a more basal sphenacodontid than Haptodus. In the large reptile tree Ianthodon was derived from a sister to Haptodus and nested at the base of Edaphosaurus + Ianthasaurus + Glaucosaurus, all edaphosaurids.

Figure 1. Ianthodon schultzei was considered a basal pelycosaur, and it is, but here nests as a basal edaphosaur. And it has no tall neural spines. So pelycosaur sails were convergent, not homologous.

Figure 1. Ianthodon schultzei (image modified from Spindler, Scott and Reisz 2015) was considered a basal pelycosaur, and it is, but here nests as a basal edaphosaur. And it has no tall neural spines. So pelycosaur sails were convergent, not homologous. Spindler, Scott and Reisz considered this specimen a juvenile due to its incomplete ossification.

Notably Ianthodon does not have tall neural spines. Earlier we wondered whether the tall neural spines of Edaphosaurus and Dimetrodon were convergent or homologous. Now it is clear, via Ianthodon, and Sphenacodon (sorry I did not notice this yesterday) that the tall neural spines of Edaphosaurus and Dimetrodon were convergent.

Most well-known pelycosaurs
were Early Permian in age. Ianthodon demonstrates an earlier origin for their carnivore/ herbivore split. And it retains carnivore teeth! Therapsids likewise originated in the Late Carboniferous according to this new data.

Phylogenetic history
Spindler, Scott and Reisz (2015) report, “In the original description and phylogenetic analysis of Kissel and Reisz (2004), Ianthodon was found to nest surprisingly high within Sphenacodontia, as a sister taxon to the clade that included Pantelosaurus, Cutleria and sphenacodontids. In a subsequent, large-scale analysis, Ianthodon was found to be more basal, near the edaphosaurid–sphenacodont node (Benson, 2012), but its exact position remained poorly resolved. In the latter analysis, Benson (2012) extensively revised the character list and included all known “pelycosaur” grade synapsids, while Kissel and Reisz (2004) used data and taxa derived from Laurin (1993), which mainly followed Reisz et al. (1992). Another recent analysis of sphenacodont synapsids by Fröbisch et al. (2011), as part of a description of a new taxon, recovered Ianthodon, Palaeohatteria and Pantelosaurus in an unresolved polygamy.”

The Spindler, Scott and Reisz (2015) analysis
used 122 characters (vs. 228 in the large reptile tree). Their tree shows 12 taxa, 4 of which are suprageneric. In their tree Ianthodon nested between Edaphosauridae and Haptodus. (So close, but no cigar.) Their tree also nested two therapsid taxa (Biarmosuchus and Dinocephalia) with Cutleria, Sphenacodon, Ctenospondylus and Dimetrodon. Thus Spindler, Scott and Reisz appear to be excluding several key taxa and their tree topology differs significantly from the large reptile tree at the base of the Therapsida, with or without Ianthodon.

References
Spindler F, Scott D. and Reisz RR 2015. New information on the cranial and postcranial anatomy of the early synapsid Ianthodon schultzei (Sphenacomorpha: Sphenacodontia), and its evolutionary significance. Fossil Record 18:17–30.

A late Jurassic pterosaur humerus from Thailand

A well-preserved pterosaur humerus
from the Late Jurassic of Thailand  has just been documented (Buffetaut et al. 2015; Fig. 1). The authors considered the humerus an azhdarchoid (traditionally tapejarids + azhdarchids, but those clades are not related to each other in the large pterosaur tree). They report, “On the basis of such an isolated specimen, a very accurate identification is hardly possible. However, the long, parallel-sided crista deltopectoralis differs from that of basal pterosaurs, which is usually shorter and broader proximodistally. The general morphology of the bone, especially its proximal region, agrees with that of azhdarchoids (sensu Witton, 2013). To sum up, the humerus from Phu Noi shows clear azhdarchoid characters, and may belong to anazhdarchid.”`

Figure 1. The new Late Jurassic Thai pterosaur humerus, PRC 64. 112 mm long.

Figure 1. The new Late Jurassic Thai pterosaur humerus, PRC 64. 112 mm long.

To test the affinities
of the new humerus to taxa in the large pterosaur tree, I first eliminated all non-pterosaurs, then eliminated all pterosaurs with a straight humeral shaft. Then I eliminated all warped deltopectoral crests. Some taxa are known from skulls only, and they were also eliminated. That left me with about 22 taxa which I then visually compared to the new Thai humerus. Two taxa were most similar (Figs. 2, 3), neither related to azhdarchids or tapejarids. One comparable (a. M. No. 4072; Fig. 2) was much too small. The other (Fig. 3), is a dorygnathid nesting at the base of the azhdarchids and protoazhdarchids (SMNS 50164; Figs. 3, 4), appeared similar to the Thai pterosaur, but occurred in Mid-Jurassic strata.

Figure 2. A tiny proto-germanodactylid, n12 in the Wellnhofer 1970 catalog is as tall as the new Thai humerus, but has a similar shape to its own humerus.

Figure 2. A tiny proto-germanodactylid, n12 in the Wellnhofer 1970 catalog is as tall as the new Thai humerus, but has a similar shape to its own humerus. Not a likely candidate, but a larger descendant is a possibility.

The tiny proto-germanodactylid, n12, a. M. No. 4072, was as tall overall as the new Thai humerus was long, so it is not a good candidate despite the similarity in numeral shapes – unless a larger version of this taxon lived in Thailand.

Figure 3. The derived Dorygnathus specimen, SMNS 50164 has a similar humerus of the right size and right age.

Figure 3. The derived Dorygnathus specimen, SMNS 50164 has a similar humerus of the right size but too early with regard to age. Not all Dorygnathus specimens have a similar humerus shape.

The derived Dorygnathus specimen (SMNS 50164) has a similar humeral shape, but occurs too early in Europe (Mid-Jurassic). However, it is possible that some dorygnathids survived in Thailand until the Late Jurassic and if so, this may be evidence of this.

Azhdarchids and Obama

Figure 3. Click to enlarge. Here’s the 6 foot 1 inch President of the USA alongside several azhdarchids and their predecessors. Most were knee high. The earliest examples were cuff high. The tallest was twice as tall as our President. This image replaces an earlier one in which a smaller specimen of Zhejiangopterus was used.

The large azhdarchids (Fig. 3) have a warped deltopectoral crest, and the small ones have a straight humeral shaft or other traits not found in the Thai humerus, so azhdarchids were dropped from consideration in the present analysis. The new Thai humerus displays several traits that were not considered in the large reptile tree. Perhaps a skull-only taxon is the closest match, but we’ll never know until more complete Late Jurassic specimens of these are found.

Also note: there is tremendous convergence within the Pterosauria. The two closest matching taxa are not related to each other. And the Thai humerus may not be related to either.

References
Buffetaut E, Suteethorn V, Suteethorn S, Deesri U and Tong H 2015. An azhdarchoid pterosaur humerus from the latest Jurassic (Phu Kradung Formation) of Phu Noi, north-eastern Thailand. Research & Knowledge 1:43-47.
Witton, M.P. 2013. Pterosaurs. Princeton University Press, Princeton and Oxford, pp. 291.

Early Permian Sail-back Synapsids

Everyone knows
about Dimetrodon and Edaphosaurus, the two Early Permian sail back synapsid reptiles (Figs. 1, 2). Ianthasaurus was a more primitive sister to Edaphosaurus. Secodontosaurus was a sister to Dimetrodon. A taxon without a sail, Haptodus, was basal to both clades.

Figure 1. Dimetrodon, a sailback pelycosaur synapsid reptile of the Early Permian.

Figure 1. Dimetrodon, a sailback pelycosaur synapsid reptile of the Early Permian.

Dimetrodon 
was a meat-eater. Edaphosaurus was a plant-eater. Every grade-schooler knows this. Skull size and sail design readily distinguish these two iconic taxa. Other traits, from teeth to toes also distinguish them.

Figure 2. Edaphosaurus, a sailback pelycosaur synapsid reptile of the Early Permian.

Figure 2. Edaphosaurus, a sailback pelycosaur synapsid reptile of the Early Permian. Note the tall caudal neural spines, distinct from Dimetrodon (figure 1).

Several specimens
of Dimetrodon are known (Fig. 3). Several attempts at reconstructing the skull of Edaphosaurus have been made (Fig. 2). I have the impression that there is not yet a single complete skull known for this taxon.

Figure 2. Click to enlarge. Sphenacodont skulls to scale. Figure 2. Click to enlarge. Sphenacodont skulls to scale.

Figure 3. Click to enlarge. Sphenacodont skulls to scale. See Figure 2 for Edaphosaurus skulls. Not sure why Sphenacodon is not considered a species of Dimetrodon. The skulls are nearly identical.

The two sails
are either convergent or homologous. At this point, we don’t know. They both have individual designs with Edaphosaurus having curved neural spines with short spars on each “mast”. If they are homologous, Ianthsaurus (Fig. 4) is close to that common ancestor. At present, sail-less Haptodus is the last common ancestor.

Figure 4. Ianthasaurus, a basal edaphosaur.

Figure 4. Ianthasaurus, a basal edaphosaur not far from the common ancestor to all tailback pelycosaurs.

Interestingly,
at the same time that sails were developing in one synapsid clade, another clade, the Therapsida, led by Cutleria and Stenocybus was developing in different ways. At present only skulls are known, but more derived therapsids had longer legs and apparently a more active lifestyle, again dividing at their origin into meat-eaters, like Biarmosuchus, and plant-eaters, like Niaftasuchus and the Dromasauria.

The Early Permian
reminds me of the Early Triassic with regard to the great amount of evolutionary novelty appearing then, likely in response to new environs, weather patterns, predators and experiments in raising the metabolism in several clades. At this time basal diapsids and basal lepidosaurs were diversifying as well.

References
Case ED 1878. Descriptions of extinct Batrachia and Reptilia from the Permian formation of Texas. Proceedings of the American Philosophical Society xvii pp. 505-530.
Cope ED 1882. Third contribution to the history of the Vertebrata of the Permian formation of Texas. Proceedings of the American Philosophical Society (20): 447–461.
Marsh OC 1878. Introduction and succession of vertebrate life in America: Popular Science Monthly, v. 12, p. 513-527, 672-697.
Modesto SP 1994. The Lower Permian Synapsid Glaucosaurus from Texas. Palaeontology 37:51-60
Reisz RR and Berman DS 1986. Ianthasaurus hardestii n. sp., a primitive edaphosaur (Reptilia, Pelycosauria) from the Upper Pennsylvanian Rock Lake Shale near Garnett, Kansas. Canadian Journal of Earth Sciences 23(1): 77–91.
Reisz R R, Berman DS and Scott D 1992. The cranial anatomy and relationships of Secodontosaurus, an unusual mammal-like reptile (Pelycosauria: Sphenacodontidae) from the early Permian of Texas. Zoological Journal of the Linnean Society 104: 127–184.
Romer, AS 1936. Studies on American Permo-Carboniferous tetrapods. Problems of Paleontology, USSR 1: 85–93.
Romer AS and Price LW 1940. Review of the Pelycosauria. Geological Society of America Special Papers 28: 1-538.

wiki/Ianthasaurus
wiki/Edaphosaurus
wiki/Secodontosaurus
wiki/Dimetrodon
wiki/Sphenacodon

Philydrosaurus: another basal choristodere

Figure 1. Philydrosaurus in several views. This specimen nests as the most basal small, short-snouted choristodere. Juveniles surround it. DGS indicates this specimen had lateral temporal fenestrae, but greater resolution may modify this hypothesis.

Figure 1. Philydrosaurus in several views. This specimen nests as the most basal small, short-snouted choristodere. Juveniles surround it. DGS indicates this specimen had lateral temporal fenestrae, but greater resolution may modify this hypothesis.

Philydrosaurus proseilus (Gao and Fox 2005, Early Cretaceous, scale bar = 2 cm) is a basal choristodere with distinct ridges on the skull over the orbits. The lateral temporal fenestra is reported as closed, but the low-resolution image provided appears to show lateral temporal fenestra. It was scored without them. The cervicals do not decrease toward the skull. Several juveniles were found associated with the presumed mother. The right skull (above) does not seem to accurately reflect the fossil. Higher resolution images have been requested.

Figure 2. Philydrosaurus compared to the BPI 2871 specimen wrongly assigned to Youngina, itself a descendant of Proterosuchus.

Figure 2. Philydrosaurus compared to the BPI 2871 specimen wrongly assigned to Youngina, itself a descendant of Proterosuchus.

Update:
The blogpost on Nundasuchus has been updated with a new reconstruction (Fig. 3) and nesting with Qianosuchus and Ticinosuchus, also reflected at reptileevolution.com.

Figure 1. from Nesbitt et al. 2014. Plus foot reconstructed here and closeups of the mandible and tooth.

Figure 3. from Nesbitt et al. 2014. Plus foot reconstructed here and closeups of the mandible and tooth.

References
Gao K-Q and Fox RC 2005. A new choristodere (Reptilia: Diapsida) from the Lower Cretaceous of western Liaoning Province, China, and phylogenetic relationships of Monjurosuchidae. Zoological Journal of the Linnean Society 145 (3): 427–444.

Astronomy vs. Paleontology

Having dealt with astronomy and paleontology for much of my life, I thought it would be a good time to compare and contrast the two.

In astronomy 
all the members of the Cosmos are available to anyone to observe with or without a telescope. All the specimens are complete with regard to their visual spectra. Interpretation is straightforward and typically not controversial.

In paleontology
all the undiscovered specimens are available to anyone who puts in the effort to find them and remove or expose them from the matrix, but some specimens cannot be excavated without a permit. Some of the discovered specimens are available for study in museums. A few discovered specimens are kept in desk drawers and offices awaiting description or redescription and are therefore unavailable. Privately held specimens cannot enter the literature, but some do. Complete specimens are relatively rare. Most to all specimens need to be reconstructed from in situ data to their in vivo state, but this is rarely done. Some bits and pieces can be misinterpreted and interpretations can be controversial. Sometimes its hard to tell a suture from a crack. Some bones are buried beneath others or leave only the faintest impressions and stains.

In astronomy
all of the visible specimens follow the law of physics and so are largely predictable and follow paradigms set down decades ago. Dark energy and dark matter remain the only enigmas. The age of the Universe and distances to various heavenly bodies appears to be universally agreed upon. Mistakes rarely if ever occur any more. No specimens need to be reconstructed: WYSIWYG.

In paleontology
most of the specimens fall readily into established clades and can be identified as to their diet and niche. However several specimens and clades have been and continue to be misidentified as to their nesting. Mistakes continue to be made largely due to taxon exclusion, sometimes by oversight, sometimes by refusal. Many determinations are made by opinion and by following tradition rather than by rigorous testing.

In astronomy
anyone can discover a member of the Cosmos, and announce it to the Astronomical Union. Time is often of the essence. The pros don’t mind if an amateur makes a discovery. Every discovery is celebrated.

In paleontology
if you discover something you have to write a paper, then submit it, then wait about six months for referees to review it, then go through the editorial process if accepted, then await its ultimate publication, often a year later. Time is never of the essence. Even so, anyone can make a contribution, if deemed acceptable, The pros don’t like amateurs making discoveries that they should be making. After all, something can only be discovered once. Some discoveries are shunned and ignored.