The myth of coding from specimens firsthand and the untapped resource of photos

You've probably heard it many times.  Advice from professional paleontologists about the proper way to code specimens.  For instance, here's Brochu from the DML in 2000-

"One thing I've noticed as associate editor of JVP is that reviewers are growing less patient with phylogenetic analyses that do not address the specimens themselves, and which instead code taxa from publications. This is being viewed increasingly as unacceptable, and I wholeheartedly embrace that view. It's the specimens that are our primary data."

I completely agree that the specimens are our primary data and that coding from specimens is preferrable to any other resource.  When I was younger back in 2000 and such, I would picture a paleontologist poring over a specimen in his hands, turning it this way and that under the light, only to triumphantly type a 0 or 1 into Nexus Data Editor and move on to the next character.  If only the world were so kind.  The dirty truth is that this is generally not the way things work, and indeed can't be, given financial and business considerations.

Any decent cladistic analysis needs a large number of taxa, and for most analyses this means specimens will be spread over the world.  For the original TWG analysis of Norell et al., seeing all the relevent specimens would mean going to the AMNH, BMNH, BPM, BSP, BYU, CEU, CMN, DINO, FMNH, GMV, HMN, IGM, IVPP, JM, LH, MNU, MOR, MUCP, NGMC, PIN, PVPH, ROM, RTMP, UA, UCMZ, USNM, WDC, YPM and ZPAL collections.  China, Mongolia, Russia, Argentina, Poland, England, Spain, Germany, Canada and over ten states of the US.  If you're lucky, you'll see the specimens on a traveling exhibit (with the caveat it usually makes them harder to examine up close) or on loan to another museum.  Many museums have casts, but these are of varying quality.  Realistically, very few paleontologists are going to have the resources to see all the specimens.  Travel cost is simply too high.

But people do manage to travel, and many papers indicate specimens were consulted for coding.  I myself visited the AMNH twice, and they happened to have many IGM specimens at the time as well.  When I write my papers, I'll put down my reference for Saurornithoides as "AMNH 6516".  But the truth is my codings don't come from looking at the specimen in person.  I saw it, I held it, sure.  But when you visit a museum collection, you get 6 hours or so per day, since they're only open for so long.  And there are usually several revelent specimens in a museum, sometimes an extremely large number (AMNH, IGM, IVPP, MOR, RTMP, etc.).  Moreover, there are usually rules about removing only one specimen from cabinets at a time, filling out cards to replace them in the meantime, etc..  And you want to be careful, since nobody wants to be "the one who dropped Ornitholestes' skull".  If I were to try to code Ornitholestes for the TWG matrix while looking at it in the AMNH collections, it would near certainly take my entire time for that day and more.  Any good matrix has at least a couple hundred characters, often several hundred.  It takes time to code.  And while people have the resources to visit museum collections, I highly doubt most have the resources to return every day for a week or two.  And realistically, matrices aren't made by having a list of characters, and running through them for every taxon, a taxon at a time.  Often comparing taxa will lead to new interpretations (as in my therizinosaur accessory trochanter example) or a taxon's morphology will lead you to redefining your states or adding a new character.  Who's going to go back to New York to see if Ornitholestes has more than ten maxillary teeth after they've rewritten their character to be "11 or more teeth" instead of "9 or more"?  And once you have a new/revised matrix several years down the line, and new taxa have been discovered, are you supposed to go on your whorlwind worldwide tour again?  Curators can do these things for specimens in their care, as can other researchers who live by a museum or have specimens on loan to them, but nobody can do them for the majority of specimens.

So how do people "code from specimens"?  They take photos.  Lots of photos.  And they code from those.  They're often better than the literature because they're in color and from as many angles as you want, but with the internet publication quality is improving.  There would be almost no reason to see Australovenator for myself, for instance, since Hocknull et al. did such a good job of photographing it.  There are certainly things photographs don't show well- sutures and restoration on some specimens, depth of depressions, some texture.  But these are hardly numerous enough to justify hundreds of dollars to see yourself.  "The literature" has gotten a bad name, but its photos can be just as good as your own, and its descriptions are usually written by people with as much or more knowledge and experience as you.  This is good news for all of us though, since it means anyone can have access to the same resources the professionals use for most specimens, without travel costs.  The internet's gone a long way to providing a Shiny Digital Future for publication access, but I think we could do more. 

What if there was an online database of specimen photos, in high resolution color, that anyone could access?  The museums' permission would be needed of course, and undescribed specimens could be excluded if under study, but it sure beats everyone spending their resources to photograph the same things.  It's also better than the current situation where people have photos of poorly described specimens, but aren't allowed to distribute them, even if they've been in the literature for over a decade and have no plans for redescription.  The odd thing is, a person is generally allowed to travel to the collection and take their own photos, but not recieve or distribute those which have already been taken.  I don't want people to think I'm just bitter about lacking access myself, as there are plenty of specimens I have photos of (both taken myself and kindly provided by others) and aren't allowed to distribute.  So I'm on both sides.  But surely such a system is broken when we're witholding information from each other that we could get for hundreds of dollars in travel fees and won't be redescribed soon anyway. 

I'd be willing to throw my (distributable) photos into such a project if someone were to set it up.  The primary obstacle besides getting museum permission would be the huge storage space, but it could probably even be done on Flickr or Picasa.  What does everyone think?

Accessory trochanters in therizinosaurs

While doing my work on the TWG matrix, I noticed something interesting.  A brief intro to the structures described here is helpful.  Theropods have an anterior trochanter (also called lesser trochanter) on the front of their proximal femur.  In birds and various maniraptoriforms it partially or completely fuses to the greater trochanter to form a trochanteric crest.  First recognized in Microvenator, a lot of taxa also have another trochanter right below the anterior trochanter, called the accessory trochanter.  Supposedly, therizinosauroids have a low, separate anterior trochanter but no accessory trochanter.  I think people have just been confusing their accessory trochanter for an anterior trochanter, while the real anterior trochanter is fused in a trochanteric crest.  Note the figure below...

The top row is from Hutchinson's (2001) femoral paper.  The anterior trochanter is 'lt' while the accessory trochanter is 'at'.  I've outlined the accessory trochanter in purple in all figures.  On the left bottom we have basal therizinosaur Falcarius (from Zanno, 2010a), which Zanno correctly realized has both an accessory trochanter and an anterior trochanter which is closely appressed to the greater.  The outline drawing is of Alxasaurus (from Russell and Dong, 1994), which was supposed to have a low cylindrical anterior trochanter.  But note that it is low in position like an accessory trochanter instead and that the greater trochanter is as wide as those taxa which incorporate the anterior trochanter into a trochanteric crest.  Next on the bottom row is an anterior view of Segnosaurus (from Zanno, 2010b).  Here too the anterior trochanter is supposed to be low (labeled flt), but again I think it matches an accessory trochanter more.  Finally, in the lower right is Chirostenotes (from Currie and Russell, 1988) in anterior view.  It was described as having a very low anterior trochanter, which would be quite unlike any avetheropod, but makes sense as an accessory trochanter.  As ornithomimosaurs and basal oviraptorosaurs have the most distinct accessory trochanters, it makes sense that therizinosaurs and Chirostenotes would too.

References- Currie and Russell, 1988. Osteology and relationships of Chirostenotes pergracilis (Saurischia, Theropoda) from the Judith River (Oldman) Formation of Alberta, Canada. Canadian Journal of Earth Sciences. 25, 972-986.
Hutchinson, 2001. The evolution of femoral osteology and soft tissues on the line to extant birds (Neornithes). Zoological Journal of the Linnean Society. 131, 169-197.
Russell and Dong, 1994. The affinities of a new theropod from the Alxa Desert, Inner Mongolia, People’s Republic of China. Canadian Journal of Earth Sciences. 30, 2107-2127.
Zanno, 2010a. Osteology of Falcarius utahensis: Characterizing the anatomy of basal therizinosaurs. Zoological Journal of the Linnaean Society. 158, 196-230.
Zanno, 2010b. A taxonomic and phylogenetic re-evaluation of Therizinosauria (Dinosauria: Maniraptora). Journal of Systematic Palaeontology. 8(4), 503-543.

Charig's thesis- "Mandasuchus" and "Teleocrater"

Fitting with the basal archosaur theme of this month, here's something I've been meaning to post for a while.  One of the rarest Mesozoic archosaur papers of all is Charig's (1957) unpublished thesis.  Unlike Zhao (1985), which I could at least order, no one can get a copy of Charig's work.  You're not even allowed to photocopy it without permission of the copyright holder, and Charig died in 1997.  Even if you somehow managed to get permission from his children, it'd still cost $176 to photocopy (assuming rates have stayed the same over the last decade, which they certainly haven't here).  George Olshevsky (DML, 2001), who's seen most every dinosaur paper in existence, believed "when he was alive he embargoed all the copies so that no one could obtain it."  In October 2001 John Jackson went to the Cambridge Library and made notes ("paraphrasing here and there, from fairly arbitrarily chosen sections, with bits missed out"), which he gave permission to distribute.  The previous person who accessed it did so in 1986, giving some idea of how rare it is.  As Charig's nomina nuda are well known, the publication is 54 years old, and he won't be publishing on them, I think the information is fair to share.  So without further ado, here's the closest any of us will probably get to reading Charig's thesis-

"Unfortunately the word “Nyasasaurus” does not appear in my notes.
 
The work is in two volumes: the illustrations: 53 plates, each with a facing
page of titles; and the volume of text of just over 500 pages.

Notes on:
Charig, A. J. 1957. New Triassic archosaurs from Tanganyika, including
Mandasuchus and Teleocrater. Dissertation Abstracts, Cambridge University.

[My comments in […] brackets.]
 
Mandasuchus longicervix, known from 3 specimens, and Teleocrater tanyura,
known from 2 specimens; in both genera, skull poor and post-cranial good.
 
General notes:
 
Both have closed acetabula.  Mandasuchus differs from Prestosuchus (from the
upper Rio do Rasto beds) in that Mandasuchus’s cervical vertebrae are
elongated (Prestosuchus’ are not, as far as is known).
 
The dermal armour differentiates the family [presumably Mandasuchus’s
family] from Stagolepidae (from the upper Tr of Laurasia).
 
There are no good grounds for considering the acetabulum of Spondylosaura[?]
to be open.  The neck of Spond. is more elongated than Mandasuchus.
 
Teleocrater has greatly elongated cervicals, unlike any other pseudosuchian
described, but closely resembling Coelophysis, yet it has a closed
acetabulum, and limbs like other pseudosuchians.
 
 
Diagnosis of Mandasuchus:
 
A pseudosuchian tending to large size.  The skull is unknown, except for
fragments of the maxilla and dentary, the former showing a large antorbital
vacuity; jaws long.  Dentition thecodont; teeth recurved, laterally somewhat
compressed, with anterior & posterior borders crenulated.
 
Vertebrae with the length of centrum never much less than the diameter and
usually greater than diameter.  Centra lightly amphicoelus; floor of neural
canal deeply concave in each centrum, except in posterior caudal region;
zygapophyses moderately oblique; tops of neural spines, especially in
anterior part of column, flattened and expanded to bear dorsal scutes.
 
Axial and caudal intercentra only.  More than 25 pre-sacral vertebrae
present.  Eight cervical vertabrae (by arbitrary definition) including
atlas; axis slightly elongated, other cervical vertebrae much elongated (up
to 50% over typical dorsals), elongation being greatest in 5th; axis and 3rd
cervical with prominent keel, faint ventromedial ridge in others; neural
spines low; axial and cervical ribs present, the later crocodiloid.
Seventeen or more dorsal vertebrae, centra mostly rounded beneath, some
slightly flattened; typical archosaurian shift in position of rib
articulation, parapophysis borne entirely on centrum in 2nd dorsal, on both
centrum and neural arch in 3rd & 4th, and on neural arch only on 5th;
diapophysis supported by oblique radiation buttress in anterior dorsals;
parapophysis and diapophysis tend to form “spectacles-shaped” rib
articulations then fuse in posterior dorsals; most if not all dorsal
vertebrae with hyposphene.  Two sacral vertebrae [in life probably].  Caudal
verts, except most anterior members, flattened ventrally and with
haemopophyses (absent in first three); distal caudals with small median
pre-neural spine between prezygapophyses, anterior to neural spine proper;
rami of proximal end of each haemopophysis joined by bridge, at least in
distal part of tail.
 
Major limb bones long and slender, with hollow shafts; propodials longer
than epipodials; bones of forelimb about two-thirds as long as hindlimb.
 
Scapula broad both dorsally and ventrally, only moderately inflected;
coracoid with small foramen; dermal elements of pectoral girdle not known.
Humerus with high deltopectoral crest, well-marked supinator process and
ectepicondylar groove, no entepicondylar foramen or groove; ulna without
olecranon; no manus.
 
Acetabulum closed.
 
Ilium with short anterior spine, long posterior spine, well-developed
supra-acetabular crest, forming most of the acetabulum; pubis long, with
small obturator foramen, twisted proximally in typical pseudosuchian manner,
distally plate-like and directed steeply downwards, thickening of lateral
corner of distal end; ischium also elongate, peduncle flattened laterally
and with sharp anteroventral edge, possibly not meeting its fellow in
mid-line but diverging from it distally, distal end slightly thickened.
 
Femur slightly sigmoidal, with prominent 4th trochanter high on shaft;
fibula with anterior muscle process; fibulare crocodyloid, pes otherwise
unknown.
 
Paramedian dorsal scutes, not corresponding in number with verts but more
numerous, keeled externally, each notched posteriorly and overlapping
anterior spine of scute behind it, without ornament.
 
 
Teleocrater:
 
First impression is a vertebral column of a saurischian type, associated
with the ilium and limb bones of a pseudosuchian.  The verts resemble a
coelurosaur very closely; long, especially anterior, cervicals, and
generally lightly constructed.  But the acetabulum certainly closed; humerus
has supinator process and ectepicondylar groove; femur has no well-defined
head set at an angle to the shaft, and is without a 4th trochanter; tibia
shorter than femur; fibula has lateral trochanter.
 
This specimen was found in a heterogeneous field-collection [but other
material easily distinguished from the Teleocrater stuff].  It seems most
unlikely that 28 verts from a saurischian, without any corresponding limb
bones, would be found together with 9 well-preserved pseudosuchian girdle
and limb bones lacking any corresponding verts, and the dimensions of the
two sets of bones being strictly comparable.
 
The genus must be referred to the pseudosuchians because of the apparently
closed acetabulum.  Neither specimen has known dermal scutes, but then the
type lacks the top of every neural spine, and the other specimen is too
incomplete.  The ilium and limb bones of Teleocrater, though typically
pseudosuchian in form, resemble no other known particularly, though not
widely different from Mandasuchus, despite great differences between the
vertebrae of the two.  On the other hand, the vertebral column *does*
resemble coelophysis.  There is a general resemblance between Teleocrater
and Coelophysis in their hollow, thin-walled bones.  Teleocrater is a little
smaller on average than Coelophysis.  In both, verts are long and slender,
especially in the neck and tail, and the centra are usually amphicoelous.
 
Von Heune’s axis [his Coelophysis one] resembles the Teleocrater vertebra in
extreme elongation and in the marked and asymmetric concavity of its ventral
profle, the apex of which lies in front of the middle of the vertebra.  It
seems that in both animals the anterior face of the centrum must have lain
more dorsally than the posterior, showing the head was held above the level
of the body.  There is further resemblance in the nature of the ventral
surface of the centrum: concave or flattened in front between a pair of
sharp edges, bearing a medial ridge in the middle part (much longer in
Coelophysis) and rounded behind.  The posterior face is roughly circular,
and fairly deeply concave in both.  The most striking similarity is between
the flange-like diapophyses and zygapophysial ridges of the two verts,
giving both a highly characteristic appearance.  The detailed arrangement of
the flanges and ridges is almost exactly the same.
..
..
..
Dorsal Vertebrae:
In both [Coelophysis and Teleocrater] still slender and rather elongate, but
a little stouter towards the sacrum.  There is also a resemblance in the
gradual change in position of [the?] parapophysi…
..
..
..
Caudal region:  There is also a general similarity here between [_T_ and
_C_].  _C_ however differs from Teleocrater in having no facets or
intervertebral spaces for haemapophyses (though Case opines chevrons would
be expected), and their ventral surfaces are rounded without traces of
longitudinal grooves or ridges.  In Teleocrater only the last 4 of the 15
caudals preserved have rounded ventral surfaces without haemapophysial
facets.  In both Teleocrater and _C_ the centra are weakly amphicoelous, the
flattened transverse processes, the bases of which are axially long in the
anterior cuadals, diminish down the series and disappear; the zygapophyses
remain, even in the smallest vert, and the anterior pair project in front of
the centrum; and the neural spines, which also diminish and disappear, have
an anterior edge rising obliquely backwards, and a posterior edge rising
almost vertically.
..
..
..
The lilum of _C_ and Teleocrater is utterly different.  In Teleocrater the
acetabulum is almost certainly closed. The anterior spine in Teleocrater is
extremely short.  The posterior spine in Teleocrater is broken off but seems
to have been long, and the beginning of a strong medial crest is preserved.
Case ’27 described an isolated ilium from Texas, similar in size to _C_
longicollis but “somewhat different in form.”  It was actually very
different, and is definitely not from _C_, but rather similar to Teleocrater
though about twice as large; any acetabular perforation must have been very
small.  Extremely long anterior spine and correspondingly long, very heavy
posterior spine, bearing a high medial crest.
 
The fumur of _C_ and Teleocrater differ mainly in the head in Teleocrater
being not at all bent towards the median side.  Teleocrater’s femur more
powerfully built but relatively even longer  (compared to _C_’s equivalent
to 5 of its dorsal verts) in Teleocrater equivalent to six or seven anterior
dorsal verts.  The shaft is sigmoidally curved and there is no ext.
trochanter.  The only similarities between femur of _C_ and Teleocrater are
the absence of a properly developed fourth trochanter, and the form of the
broadened, club-shaped distal end with its weakly developed condyles.
 
Tibia:
Not the remotest resemblance between the two animals.
 
Fibula: in _C_ “but little expanded.”  This differs from Teleocrater.
 
Summary:
_C_ and Teleocrater only similar in their vertebral columns.  This
similarity is no more than might be expected between any two coelurosaurs.
However there is a particularly close resemblance in the anterior cervicals,
with their highly elongate centra and characteristic flanges; this form is
most unusual and yet almost identical.
..
..
..
General Taxonomics:
Both Mandasuchus and Teleocrater are from the “upper bone bed” of
Tanganyika, and share a common tendency to resemble the saurischia to an
unusual extent.  This tendency, particularly in Teleocrater, is more
pronounced in the vertebral column than the girdle and limbs.  The two
genera are strikingly different: Mandasuchus (and other prestosuchids)
generally reminiscient of pachypodosaurs, Teleocrater very like
coelurosaurs.
 
[He considers the vert. column similarities between Teleocrater and
Mandasuchus enough to suggest acet. perforation occurred independently in
the two saurischian groups (and between 4-6 times overall).]
 
 
 
His thesis also deals with other archosaur bits from the Maleri beds of
India, all of which were described in Von Heune (‘40c): Palaeont. Indica
(n.s.) 32 memoir no. 1, 1-42.
 
 
From the Summary:
The only important difference between Mandasuchus and Prestosuchus is
Mandasuchus’s cervical verts are elongated while the latter’s (as far as is
known) are not.  It is proposed to erect a new family: Prestosuchidae for th
e two, the most diagnostic feature of which would be the dermal armour,
clearly distinguishing it from the Stagonolepidae." 

Thanks to John for this interesting look at an obscure document.

Is Saltopus a silesaurid?

Saltopus was recently redescribed by Benton and Walker (2011), who included it in Brusatte et al.'s (2010) archosaurian matrix.  The result is that Saltopus is a dinosauriform more derived than Pseudolagosuchus but outside Silesauridae(including Lewisuchus)+Dinosauria.  Since many recent analyses have synonymized Pseudolagosuchus and Lewisuchus, Tim Williams asked on the DML whether Saltopus would also be a silesaurid if Pseudolagosuchus and Lewisuchus were combined in Brusatte et al.'s matrix.  Thanks to Tim, I have the Saltopus paper, so now the question can be answered.

First, I made a NEXUS file from Brusatte et al.'s matrix.  Characters 52, 156, 159 and 170 had to have their states switched so that they could be properly ordered.  More characters were ordered than Brusatte et al. had, with only 14, 51 and 96 left unordered.  I then made a new OTU combining Lewisuchus' and Pseudolagosuchus' codings.  Only two codings differed.  Character 134 is the ventral emargination of the femoral head in anterolateral view seen in lagerpetonids.  This was scored as present in Lewisuchus as well, and while Romer's (1972) figure does seem to show it, Arcucci (1998) determined this is actually a tibia.  The other differing character is 148, the tibiofemoral ratio, which again is caused by the misidentified tibia.  Thus the femoral codings of Lewisuchus were not taken into account for the combined OTU.  The resulting tree is below-

|--Erythrosuchus
`--+--Euparkeria
   `--+--Proterochampsidae
      |--+--+--Paleorhinus
      |  |  `--+--Mystriosuchus
      |  |     `--Rutiodon
      |  `--+*-Stagonosuchus
      |     |*-Tikisuchus
      |     |--+--Qianosuchus
      |     |  `--+--+--Aetosaurus
      |     |     |  `--+--Desmatosuchus
      |     |     |     `--Stagonolepis
      |     |     `--+--Gracilisuchus
      |     |        `--+--Erpetosuchus
      |     |           `--+--Protosuchus
      |     |              `--+--Sphenosuchus
      |     |                 `--Terrestrisuchus
      |     `--+--+--Revueltosaurus
      |        |  `--+--Ornithosuchus
      |        |     `--Riojasuchus
      |        `--+--+--Rauisuchus
      |           |  |--Saurosuchus
      |           |  |--+--Batrachotomus
      |           |  |  `--Prestosuchus
      |           |  `--+--Postosuchus
      |           |     `--Teratosaurus
      |           `--+--+--Arganasuchus
      |              |  `--+--Fasolasuchus
      |              |     `--Ticinosuchus
      |              `--+--Yarasuchus
      |                 `--+--+--Lotosaurus
      |                    |  `--+--Effigia
      |                    |     `--Shuvosaurus
      |                    `--+--+--Arizinasaurus
      |                       |  `--Bromsgroveia
      |                       `--+--Poposaurus
      |                          `--Sillosuchus
      `--+--+--Scleromochlus
         |  `--+--Dimorphodon
         |     `--+--Eudimorphodon
         |        `--Pteranodon
         `--+--+--Dromomeron
            |  `--Lagerpeton
            `--+--Marasuchus
               `--+--Lewisuchus (inc. Pseudolagosuchus)
                  `--+--Saltopus
                     `--+--+--Silesaurus
                        |  `--+--Eucoelophysis
                        |     `--Sacisaurus
                        `--+--+--Psittacosaurus
                           |  `--+--Heterodontosaurus
                           |     `--Lesothosaurus
                           `--+--Herrerasaurus
                              `--+--Plateosaurus
                                 `--Coelophysis

Note that basically, the combined Lewisuchus+Pseudolagosuchus goes where Pseudolagosuchus used to.  So it really doesn't answer our question.  Also note that proper ordering changed a lot of relationships within Suchia from Brusatte et al.'s tree.

To resolve whether Saltopus would be a silesaurid in a topology where Lewisuchus is a member of the family, I constrained the results to agree with Nesbitt's tree for all taxa used in both analyses.  The result is 47 steps longer and is interesting in-

- finding Gracilisuchus sister to Revueltosaurus+Aetosauria (as in some trees of Nesbitt's).

- finding Erpetosuchus sister to Crocodylomorpha (as in Brusatte et al.).

- finding Arganasuchus to be a pseudosuchian outside Ornithosuchidae, Gracilisuchus+Aetosauria, Poposauroidea and Teratosauridae+Crocodylomorpha.

- finding Stagonosuchus to be the most basal loricatan.

- finding Tikisuchus to be a teratosaurid (rauisuchid of Nesbitt) (as in Brusatte et al.).

- finding Yarasuchus the most basal poposauroid (as in Brusatte et al.).

- finding Bromsgroveia to be at least as derived as Arizonasaurus but outside Lotosaurus+Shuvosauridae (as in Brusatte et al.).

- finding Scleromochlus to be a pterosauromorph (as in Brusatte et al.).

Saltopus forms a polytomy with Dinosauria and all four silesaurids, so may be a silesaurid or may not.   And we're no closer to answering the question that when we began.  It might be more useful to code it for Nesbitt's matrix.

Where are the pterosaurs recoded for Nesbitt's matrix, you may ask.  I'm afraid once again work on a paper to be published has superceded this more recreational study.  Luckily, it's work that coincides with my TWG analysis and should result in coauthorship of a new taxon. I will say that Nesbitt didn't code several pectoral and forelimb characters for pterosaurs, which when coded make them less likely to be archosaurs.  However, the braincase of Pteranodon makes them more likely to be archosaurs.  So after adding Preondactylus and Pteranodon, and recoding Eudimorphodon and Dimorphodon (but with no hindlimb characters looked at yet), they're still avemetatarsalians and take 17 steps to move them outside Archosauria by Peters' 'key taxa'.

Sharovipteryx in Nesbitt's matrix

The last of Peters' (2000) "key taxa" to placing pterosaurs outside Archosauriformes is Sharovipteryx.  Another taxon which was originally poorly illustrated and described, Gans et al. (1987) did a mediocre job redescribing it, and good photos are available of some areas.  Unwin et al. (2000) provide some new information.  Like Longisquama, Peters' imagination is at play in the 2000 paper.  The forelimbs were not exposed at the time and the skull is in dorsal view, not ventral.

After coding, Sharovipteryx joins the other discussed taxa somewhere basal to Erythrosuchus.  The topology is- (Sharov(Cose,Lango)(Longi(Valle,Megalanc))).  As before, 13 more steps are needed to take pterosaurs out of Archosauriformes.  They're still closest to simiosaurs, then Longisquama, with Sharovipteryx the next outgroup.

Next up, the pterosaurs themselves.

Simiosaurs in Nesbitt's matrix

Simiosaurs such as Megalancosaurus are one of the most controversial groups of diapsids.  While generally placed among basal archosauromorphs, they've also been considered non-saurians (Senter, 2004), euryapsids (Muller, 2004), lepidosauromorphs (Peters, 2007) and theropods (Olshevsky, 1991).  They've also been attached to both pterosaur (Peters, 2000; Renesto and Binelli, 2006) and bird (Feduccia and Wild, 1993) ancestry. 

Enforcing Cosesaurus and Longisquama to be pterosauromorphs is five steps longer, and they along with simiosaurs and Langobardisaurus end up as basal avemetatarsalians.  Enforcing Cosesaurus and Longisquama to be pterosauromorphs and placing the clade outside Archosauriformes is thirteen steps longer.  In this topology, they are prolacertiforms, simiosaurs and closest to pterosaurs, with Longisquama one node out.  This is ten steps longer than Renesto and Binelli's (2006) topology where simiosaurs+pterosaurs are sister to archosauriforms and prolacertiforms with Langobardisaurus and more basal.  Enforcing Megalancosaurus to be a theropod is 34 steps longer (Longisquama and Vallesaurus follow it).  I neglected to mention last time that Olshevsky also placed Cosesaurus in Theropoda, which takes 41 more steps, with simiosaurs, Longisquama and Langobardisaurus following.  So Olshevsky was near certainly wrong about either Megalancosaurus or Cosesaurus being theropods, though Longisquama is more equivocal.

Notice the trend for how much less parsimonious it is to contrain pterosaurs as non-archosauriforms.
- Longisquama added. 23 more steps.
- Longisquama and Cosesaurus added. 22 more steps.
- Longisquama, Cosesaurus and Langobardisaurus added. 15 more steps.
- Longisquama, Cosesaurus, Langobardisaurus and simiosaurs added. 13 more steps.

Next up before I add Preondactylus and Pteranodon (and check the other pterosaurs' coding) is Peters' (2000) last "key taxon" Sharovipteryx.  Will it make non-archosauriform pterosaurs even more parsimonious?

Langobardisaurus in Nesbitt's matrix

Continuing the series, I added the tanystropheid Langobardisaurus to Nesbitt's new archosauriform analysis.  Langobardisaurus is another taxon which Peters (2000) considered key to connecting prolacertiforms with pterosaurs.  It's very odd in its own right, especially its somewhat artiodactyl-like skull.  Luckily, several specimens have been well described and it's better preserved than Longisquama or Cosesaurus.  Once coded, it groups with Longisquama and Cosesaurus, but this clade can vary in position.  It can be anywhere outside Pseudosuchia or Ornithodira.  Constraining prolacertiform pterosaurs is now only 15 steps longer, compared to 22 steps before Langobardisaurus was added.  What will happen when Vallesaurus and Megalancosaurus are added?

Cosesaurus in Nesbitt's matrix

I haven't written the Cosesaurus entry for my site yet, but to me it always seemed the BANDits best bet, though instead they went with the weird Megalancosaurus and parafeathered Longisquama.  Cosesaurus looks outwardly like a little theropod, though of course it differs in the details.  Like Longisquama, its anatomy has not been redescribed since the 70s and the existing reconstructions involve lots of imagination.  The thing's preserved as a sandstone impression, and Ellenberger (1977) saw lots of feathers and drew numerous lines in and around the skull that have no guarantee of reflecting morphology.  Peters (2000) handled it better, but like in Longisquama identified antorbital fenestrae and suture details that Ellenberger didn't see, which makes me suspicious.  So again I coded it conservatively, using only the sutures agreed upon by both authors and Peters' interpretation of the pelvis.  I'm curious about including Cosesaurus not only because it's one of Peters' proposed non-archosaurian pterosaur relatives but also to address the argument that pterosaurs only clade with avemetatarsalians because there's nothing better in the matrix.  Would Cosesaurus also clade with avemetatarsalians despite its primitive anatomy?  Would it draw pterosaurs out of Archosauria?

The answer is that Cosesaurus emerges basal to Erythrosuchus, in a polytomy with Mesosuchus, Prolacerta, proterosuchids and higher archosauriforms.  If Longisquama is included as well, both clade with Prolacerta.  So it seems Nesbitt's matrix handles basal taxa just fine.  Only four steps move it sister to pterosaurs in Avemetatarsalia though, which is somewhat worrying.  It takes 22 more steps to move pterosaurs out with them in Prolacertiformes.

We'll see how this develops with the addition of unambiguous tanystropheid Langobardisaurus, a couple simiosaurs, Preondactylus and Pteranodon.  I'll also be checking the general pterosaur codings when coding them, to make sure Nesbitt didn't assume homology a priori by e.g. coding manual digit IV as unknown in amount of reduction.

Longisquama in Nesbitt's matrix

Here's an interesting one.  Longisquama has an extremely controversial phylogenetic position.  It's been placed outside Sauria, as a lepidosauromorph, a tanystropheid, a non-dinosaurian archosauromorph and even a theropod.  See my discussion here for details.  Nesbitt's matrix cannot test most of these, since it only covers archosauriforms.  But I was curious what would happen if it was added to the matrix, since it's been allied with pterosaurs by some, and the BAND crowd insists it's an archosauriform.  Not that the latter would believe a cladistic analysis, but meh...

Unfortunately, Longisquama's morphology is also controversial, and has yet to be studied in depth.  Such basic features as the antorbital fenestra and tooth implantation are not agreed on by all authors.  So I coded Longisquama twice.  One uses Peters' (2000) interpretation of skull elements and the sternum-interclavicle structure.  The other uses the minimum codable without dividing the skull into discreet elements and treating the aforementioned pectoral structure as "bony wisps that defy interpretation" as interpreted by Senter (2003).

Perhaps unsurprisingly, its position depends on how it's coded.  Coding it conservatively results in Longisquama being outside Archosauriformes alongside Mesosuchus and Prolacerta.  Constraining it and pterosaurs to be sister taxa is only two steps longer, and puts both in Avemetatarsalia.  Constraining it as a theropod is seven steps longer, and it emerges as the basalmost one.  Coding Longisquama as Peters illustrates it (this was before he thought he saw the hindlimbs and such) results in it being a pterosauromorph, with pterosaurs in their usual position.  Constraining it to be outside Archosauriformes is only three steps longer.

In conclusion, Longisquama desperately needs redescribed.  Ignore the parafeathers, we need its basic anatomy established.  While moderately rejected as a theropod, I'd say it could equally well be a non-archosauriform or a pterosauromorph regardless whose reconstruction is used.

Lukousaurus in Nesbitt's matrix

Back in ye olde days when I was new to the DML, I wrote a post on Lukousaurus to determine its phylogenetic affinities- http://dml.cmnh.org/2000Sep/msg00086.html .  I didn't have anywhere near the amount of knowledge and resources I do now, and I considered it an abelisaurid or sphenosuchian.  Times have changed and now we have Nesbitt's huge new analysis which is perfect for coding Lukousaurus in.  It doesn't have abelisaurids, but does have Dilophosaurus and Allosaurus along with several sphenosuchians.  Note the caveat that Lukousaurus is coded from Young's 1948 figures and description.  So after entering Lukousaurus, there are 4683 MPTs with Lukousaurus as an archosaur in all of them.  It is not a member of Ornithosuchidae, Revueltosaurus+Aetosauria, Poposauroidea, Loricata or Dinosauromorpha.  So it's probably some sort of basal pseudosuchian...

By the way, I noticed when coding that as in Smith et al.'s matrix, Marasuchus is not coded for its maxilla, even though Nesbitt lists the specimen as one of those he used for scoring, and lists the maxilla's presence in the material.  Also Coelophysis is oddly coded for osteoderm character 408, despite lacking osteoderms.

More added taxa to follow...

Nesbitt's excellent new analysis

I just wanted to state my praise for Sterling Nesbitt and his amazing new archosaurian phylogeny paper.  The taxon sampling is extensive, with discussions of which material is used for each taxon.  The characters are described in great detail and illustrated wonderfully, with the specimens they're found in often listed.  Alternative phylogenies are examined and character support is discussed.  The matrix seems completely coded.  I chuckled at the dismissals of Lucas' "superficial" taxonomic discussions.  :)  This is surely one of the best archosaurian analyses published, and that includes ALL archosaurian analyses, on theropods, birds, crocs, you name it.  But it just wouldn't be like me to not include some criticisms, so I offer the following...

On page 236, Nesbitt discusses the position of pterosaurs, which is as basal avemetatarsalians.  Note this doesn't mean much in regard to the controversy of their wider relationships among reptiles, since simiosaurs, Longisquama and tanystropheids were not included.  Nesbitt tries to counter Bennett's (1996) statement that pterosaurs don't have many synapomorphies of archosauriform clades more inclusive than Ornithodira with a list of the synapomorphies of Crurotarsi sensu lato, Archosauria, Archosauriformes, etc., detailing which are found in pterosaurs.  This is all fine except that 13 of the characters are simply stated to be "not known in basal pterosaurs".  I can't help but wondering if these (generally difficult to observe features in the braincase, tarsals and such) are known to be absent in derived pterosaurs.  If so, they should still count against pterosaurs being in the listed clades.  In fact, I wonder why no derived pterosaur like Pteranodon was included, when derived members of other clades (e.g. Alligator, Allosaurus, Velociraptor) were used to increase the informativeness of characters.

Note Nesbitt's Rauisuchidae should be called Teratosauridae, assuming Teratosaurus is in its standard position by Polonosuchus and Postosuchus.

I would have liked to see Doswellia, Trialestes, "Zanclodon" arenaceus, Erpetosuchus (oddly included in the character descriptions, but not the matrix), Hallopus and/or Macelognathus, Scleromochlus and Preondactylus included.

On page 187, Nesbitt is right to note that many analyses have "Characters just listed with no, little, or
vague explanations".  How reduced is "reduced"?  How tapered is "tapered"?  Unfortunately, Nesbitt's characters sometimes have the same problem.  For instance, character 282 is "Pubis, length: (0) shorter or subequal to the ischium; (1) longer than ischium."  The word "subequal" should be expunged from character descriptions, since it's not quantified.  If I have a taxon with a puboischial ratio of 101%, that's presumably subequal.  But what if it's 103%?  106%?  Everyone's conception will be different, and this will lead to miscodings.  Or character 259- "Metacarpal IV: (0) present; (1) reduced to a nubbin or absent."  When is a metacarpal a "nubbin"?  Or character 266- "Ilium, crest dorsal to the supraacetabular crest/rim: (0) vertical; (1) anterodorsally inclined."  Vertical compared to what basis of horizontal (sacrum axis, axis between anterior and posterior tips, axis formed by ventral extent of peduncles, etc.), and would 92 degrees count as vertical or inclined?  These kinds of issues are found in almost every analysis, so this isn't a critique of Nesbitt as much as it's an urging to all of us to explicitly define our character states.

But these issues aside, the paper is extremely well done.  Besides the above taxa, someone should use this matrix to place Arganasuchus, "Cryptoraptor", Lagosuchus, Lukousaurus, Razanandrongobe, Saltopus, Spondylosoma, Stagonosuchus, Yarasuchus and Zanclodon. 

Nesbitt, 2011. The early evolution of archosaurs: Relationships and the origin of major clades. Bulletin of the American Museum of Natural History. 352, 292 pp.

Pneumatocrania and Holtz's thesis

Everyone knows Holtz's famous 1994 paper revolutionizing theropod phylogeny.  What fewer people have seen is the 1992 thesis that preceded it, which included its own phylogenetic analysis.  The analysis had largely similar characters (128 vs. 126) and included Procompsognathus, Piatnitzkysaurus, Eustreptospondylus and Chilantaisaurus in addition to the taxa from 1994.  It also split Ornithomimosauria into Harpymimus, Garudimimus and Ornithomimidae, and did not include Abelisauridae.

The results are similar to the 1994 paper, with the following differences.
- Procompsognathus is outside Eutheropoda.
- Elaphrosaurus is a non-avetheropod tetanurine more derived than Torvosaurus, though on page 264 he notes the characteristics of abelisaurids indicates they and Elaphrosaurus are neoceratosaurs.
- Megalosaurus is a basal carnosaur (along with Piatnitzkysaurus).
- Chilantaisaurus is carnosaur in a trichotomy with Allosaurus and Acrocanthosaurus, and Eustreptospondylus is sister to this group.

Some new names are proposed.
- Eutheropoda is the group containing Ceratosauria sensu lato and Tetanurae, but not Procompsognathus.  It's been used informally this way online since the 90s, including on Holtz's site, but I think Novas et al. (2003) have been the only authors to publish it.  They used it in the same way (excluding Herrerasaurus and Eoraptor this time), but it's just an abstract.
- Eumaniraptora is first used here, though in a different sense than what was later adopted once it was published in 1997.  In Holtz's thesis, Eumaniraptora is what we would now call Maniraptoriformes- the group containing paravians, oviraptorids and arctometatarsalian taxa.
- Holtz used Dinoaves for the Paraves group of dromaeosaurids and Archaeopteryx, though this originally referred to something like Tetanurae when Bakker coined it.
- Arctometatarsalia is proposed here, with the same content as the 1994 paper.  
- Pneumatocrania is a name that hasn't made it to the internet before now.  It was Holtz's name for his clade of oviraptorids and 'arctometatarsalians', since they were diagnosed in part by pneumatic skulls.  Of course, we now know dromaeosaurines reduced their pneumaticity and most of the characters used are problematic (see characters 17, 64, 68, 69, 103, 115, 116 and 121 in my link above), so the topology hasn't been found since.

My favorite part of the thesis is the detailed discussion of twenty-nine poorly known theropod taxa.  This includes prescient aspects such as a maniraptoriform Microvenator and a suggestion of elaphrosaur affinity for Chuandongocoelurus almost a decade before I thought the same thing on the DML.  It also features some common but incorrect ideas of the early 90's though, such as neoceratosaurian spinosaurids and Carcharodontosaurus.

Thanks to Tom Holtz for permission to write about this interesting piece of theropod history.

References- Holtz, 1992. An unusual structure of the metatarsus of Theropoda (Archosauria: Dinosauria: Saurischia) of the Cretaceous. Unpublished PhD thesis. Yale University. 347 pp.

Holtz, 1994. The phylogenetic position of the Tyrannosauridae: Implications for theropod systematics. Journal of Paleontology. 68(5), 1100-1117.

Novas, Haro and Canale, 2003. Un nuevo terópodo basal de la Formación Ischigualasto (Carniano) de la provincia de San Juan, Argentina. Ameghiniana. 40(4), 63R.