Thursday, April 25, 2013

Elmisaurus elegans no more? Leptorhynchos

Jaime Headden, Lord and Master of All Oviraptorosaurian, just wrote a detailed commentary on this work, but my commentary is limited to the phylogenetic analysis and Elmisaurus monophyly.  Just so everyone's up to speed, Longrich et al. (2013) describe some new caenagnathid remains from Texas, revise the taxonomy of the whole clade, and present a phylogenetic analysis of oviraptorosaurs.  Among their actions is to create the genus Leptorhynchos for Elmisaurus elegans and their new species gaddisi

The analysis in the paper is represented as an Adams consensus that leaves out two included oviraptorosaurs- Epichirostenotes and Nomingia.  For those of you who don't know, Adams consensus' can be misleading.  What they do is move unstable taxa to the base of the smallest clade they definitely belong to.  So if e.g. segnosaurs were most parsimoniously either plateosaurs like Sereno thought or oviraptorosaurs like Russell and Dong thought, an Adams consensus would move them to the base of Saurischia where no one was advocating they go.  Which is fine in an incertae sedis sort of way, except that PAUP doesn't tell you when any taxon actually is basal in a clade or is just being displayed basally because it can occupy multiple more derived positions.  Also, Longrich et al. excluded Nomingia and Epichirostenotes a priori instead of a posteriori, which meant none of their information affected the topology.

Analyzing every taxon, but only leaving those in which preserve mandibles results in-
(Microv(Gigant(Caenagnathasia(elegans,gaddisi(pergracilis(collinsi,sp.))))))

The Adams consensus places Elmisaurus rarus out by Caenagnathasia, making it seem like the analysis supports creating a new genus for elegans.  But that's just an illusion caused by E. rarus not preserving mandibles, so it can go many places in Caenagnathidae and is placed basally due to the Adams consensus artifact noted above.  Indeed, if we delete taxa which don't preserve feet a posteriori, we get-

(Microv(Gigant(elegans,pergracilis,rarus)))

So it seems rarus isn't separated from elegans. Trees constrained to place elegans in Elmisaurus are equal length to trees which don't.  How many Elmisaurus apomorphies were included?

1. tarsometatarsal fusion- included, but E. rarus miscoded as lacking metatarsal fusion.

2. posterior surface of metatarsus deeply concave- not included.

3. metatarsals II and IV subequal in length to III (>93%)- not included.

4. distal end of metatarsal II curves anteriorly away from the longitudinal axis of the metatarsus- not included.

5. proximolateral process on metatarsal IV- not included.

6. more extensive proximodistal contact between metatarsals II and IV- not included.

Longrich et al. do claim elegans is more similar to Chirostenotes in one way though.  They state "Longrich (2008a) tentatively placed [elegans] in Chirostenotes, because the third metatarsal has an anteroposteriorly flattened shaft that is concave and broadly exposed on the posterior of the metatarsus (Currie 1989). This is a derived feature found in Chirostenotes (Currie and Russell 1988) but not Elmisaurus (Osmólska 1981)."  This seems related to their new character 205- "Metatarsal III with an ovoid or subtriangular cross section (0) or anteroposteriorly flattened, with a concave posterior surface (1). Primitively in theropods the third metatarsal has an ovoid cross section, or a triangular cross section in arctometatarsalian forms. This condition is retained in most oviraptorosaurs, including the basal caenagnathid Elmisaurus rarus. In Caenagnathinae, the third metatarsal is anteroposteriorly compressed."

Posterior view of proximal left caenagnathid tarsometatarsi, metatarsal III colored blue.  Left- Elmisaurus/Leptorhynchos elegans holotype ROM 781 (after Currie, 1989); Middle- Elmisaurus rarus holotype ZPAL MgD-I/172 (after Osmolska, 1981); Right- Elmisaurus/Leptorhynchos elegans referred RTMP 92.39.4 (after Currie, 1989).

Yet the posterior transverse exposure of proximal metatarsal III proximally seems intermediate in E. rarus' holotype compared to the two specimens of elegans (see figure above).  More distally, Currie's (1989) figure 2P section indicates the posterior exposure is narrow as in E. rarus (see lower left section below).  Currie's (1989) figure 2Q shows E. rarus has a concave posterior metatarsal III surface as well (see middle left section below, outlined in blue). 

Left- cross sections of caenagnathid metatarsi after Currie (1989), anterior/extensor to top, anterior and posterior surfaces lined with red and blue respectively; Top- Elmisaurus/Leptorhynchos elegans referred RTMP 92.39.4; Middle- Elmisaurus rarus holotype ZPAL MgD-I/172; Bottom- Elmisaurus/Leptorhynchos elegans holotype ROM 781.  Green lines indicate region of sectioning.  Right- Holotype of Elmisaurus rarus in anterior (left) and posterior (right) views (after Olsmolska, 1981).

As for shape, the main issue seems to be we're dealing with different proximodistal points along the bone.  Currie and Russell (1988) state in Chirostenotes pergracilis "The proximal end, viewed dorsally, is diamond shaped, tapering both anteriorly (between the contact of metatarsals 11 and IV) and posteriorly. Its major horizontal axis, 17.5 mm long, is anteroposterior in orientation and thins backwards."  Sternberg (1932) also states the Macrophalangia holotype (possibly Caenagnathus collinsi according to Longrich et al.) has a transversely compressed proximal end.  The proximal ends of E. rarus' and elegans' metatarsal III are fused too well with surrounding bones to compare.  Once C. pergracilis' anterior surface is exposed, "the bone twists until the medial surface is facing anteriorly" and "the anterior edge has broadened out to 7.5 mm to separate the adjacent metatarsals and is triangular in section" (Currie and Russell, 1988).  elegans' holotype is broken at about this same point and also shows a triangular section (upper section left above).  Note a triangular section is what Longrich et al. are claiming caenagnathines don't have, though it exists in both pergracilis and elegans.  Currie (1989) was also wrong in comparing the proximal diamond shape of Chirostenotes pergracilis with the more distal triangular shape of elegans, and none of these areas have been described in E. rarus (thus I need to fix that on the Database).  At two-thirds down in Chirostenotes, "In section, a shallow concave surface faces posteriorly at this level, while slightly concave surfaces face posteromedially and posterolaterally for contact with the adjacent metatarsals."  This appears similar to E. rarus from what the anterior and posterior views suggest (right above), and matches how Snively (2000) described an RTMP "Elmisaurus sp." metatarsus that is probably elegans.  It also matches the cross section of elegans illustrated by Currie (1989- fig. 2P; lower left section above).  While Currie's figure 2Q of Elmisaurus rarus (middle left section above) would suggest a slightly different shape where the articular surfaces are smaller and that for metatarsal IV doesn't angle posteromedially, the narrow anterior exposure of metatarsal III means it must have been taken more proximally, probably about halfway down considering the ratio between anterior and posterior exposure of metatarsal III.  This leaves anteroposterior compression, which varies throughout the bone in elegans at least.  Distally it's transversely compressed but proximally it's anteroposteriorly compressed.  The E. rarus section which is probably intermediate in position is also intermediate in compression, being slightly transversely compressed.  If Chirostenotes/Caenagnathus specimens are any indication, the bone switches back to transversely compressed at its proximal tip.  So there are actually no valid described differences in metatarsal III sectional shape between Chirostenotes and any Elmisaurus species in the primary literature.

Correcting the coding for Elmisaurus rarus having fused metatarsals and adding the 5 other characters (one as an additional ordered state of the arctometatarsaly character) results in elegans and gaddisi being in a trichotomy with E. rarus.  This elmisaur clade is sister to a caenagnath clade containing Chirostenotes, Hagryphus, Caenagnathus collinsi and C. sp. from Hell Creek.  Nomingia and Epichirostenotes both have uncertain positions in the elmisaur-caenagnath clade, which is sister to Caenagnathasia, then Gigantoraptor as in Longrich et al.'s trees.

Given this result, I wonder why Longrich et al. didn't include the Elmisaurus characters suggested by Currie.  Also, making elegans the type of Leptorhynchos puts us in a confusing situation, as the only elements described in all three species are distal metatarsal III and proximal metatarsal IV.  Neither of these seem particularly distinctive in any of the species, and rarus' limited illustrations and description makes them difficult to compare in depth.  Thus there's no obvious reason to refer gaddisi to either Elmisaurus or Leptorhynchos, so it would have been easier and more accurate to just keep elegans in Elmisaurus and not name a new genus.  Longrich et al.'s diagnoses for Leptorhynchos and each of its species are purely mandibular, so we can't even evaluate Elmisaurus rarus or the elegans holotype for them.  But if elegans' diagnosis is based only on characters not observable in the holotype metatarsus, then doesn't that make the species indeterminate?  If the holotype can't actually be distinguished from the two metatarsal fragments referred to gaddisi, we have a problem.  Luckily, it seems Longrich et al. never assigned a type species to Leptorhynchos, making it unavailable (thanks to Cay on Jaime's blog).  I really think it should just be dropped and gaddisi assigned to Elmisaurus, though even then it might only be distinguishable from referred specimens of E. elegans, and perhaps not at all from E. rarus.  Ugh, caenagnathids.

References- Osmolska, 1981. Coossified tarsometatarsi in theropod dinosaurs and their bearing on the problem of bird origins. Palaeontologia Polonica. 42, 79-95.

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.

Currie, 1989. The first records of Elmisaurus (Saurischia, Theropoda) from North America. Canadian Journal of Earth Sciences. 26, 1319-1324.

Snively, 2000. Functional morphology of the tyrannosaund arctometatarsus. Unpublished Masters Thesis. 273 pp.

Longrich, Barnes, Clark and Millar, 2013. Caenagnathidae from the Upper Campanian Aguja Formation of west Texas, and a revision of the Caenagnathinae. Bulletin of the Peabody Museum of Natural History. 54(1), 23-49.

Friday, April 19, 2013

Are Daemonosaurus, silesaurids and poposaurs phytodinosaurs? plus conclusion on Peters

No, of course not.  Phytodinosauria doesn't even exist.   But Peters doesn't believe that, due to the many issues discovered in my evaluation of his analysis.  Last time we looked at the characters he uses to support his heterodox Theropoda.  This time we'll look at the characters he uses to support his heterodox Phytodinosauria and internal clades.  A spoiler for those who don't want to read further- none of Peters' clades are even plausible except for ornithischian Daemonosaurus, which is rejected by "only" 7-10 more steps compared to 25-76 more steps for his other clades.

Phytodinosauria (Panphagia, Pampadromaeus, Sauropodomorpha, Daemonosaurus, Ornithischia, Silesauridae, Poposauridae)
8. Posterior maxilla not flared laterally to form constricted snout.  This was coded by Peters for way too many taxa, and actually has a very mixed distribution.  Effigia, Coelophysis (miscoded), Pantydraco and Heterodontosaurus (not coded) lack it, while Lotosaurus (miscoded), Herrerasaurus, possibly Pisanosaurus (not coded) and Lesothosaurus (miscoded) have it.  Daemonosaurus cannot be coded, contra Peters.
22. External naris faces dorsolaterally instead of laterally.  Peters again codes this for many more taxa than can be evaluated, and miscodes Heterodontosaurus and Silesaurus as dorsolateral when they are actually lateral.  In Peters' topology it's still recovered as supporting this clade, but barely.  Daemonosaurus cannot be coded, contra Peters.
45. Frontal without posterior process.  This was an example in my list of characters misunderstood by Peters.  No included taxon has such a process.
69. Squamosal and quadratojugal form a semicircular indentation in the infratemporal fenestra.  Peters codes this as present in Pampadromaeus (actually lacks it), Lotosaurus, Silesaurus (actually unknown) and Shuvosaurus (lacks it).  That already disqualifies it, but I'll also note that the not included Plateosaurus and Massospondylus carinatus can have a semicircular or pointed indentation, so it's quite a variable character.  Again Daemonosaurus cannot be coded, contra Peters.
119. Dentary decurved.  This is miscoded in Panphagia, Massospondylus kaalae and Daemonosaurus, and not coded in Saturnalia (actually absent).  So only Pampadromaeus and Pantydraco are correctly coded as having it, and I have no issue with this being a synapomorphy within Sauropodomorpha.
125. Posterior mandible deepest anteriorly.  This was miscoded as present in Pampadromaeus, Panphagia, Shuvosaurus (actually unknown) and Effigia.  So it would actually diagnose the sauropodomorph+ornithischian node in Peters tree, though its absence in the non-included Efraasia makes its convergent development in each group likely.  Again, this is miscoded in Daemonosaurus.
Results-  Of six characters, Daemonosaurus has one (which is found in all examined taxa) and lacks two.  Only the external naris being angled more dorsally even potentially supports the clade.

(Sauropodomorpha, Daemonosaurus, Ornithischia, Silesauridae, Poposauridae)
6. Premaxilla a third or more of preorbital length.  This is miscoded in Pantydraco and Silesaurus.  It actually distinguishes ornithischians and poposaurids, while also being present in Massospondylus kaalae (but not the non-included M. carinatus, Plateosaurus or Efraasia).  Yet again, it is miscoded in Daemonosaurus.
11. Nasals widest at midpoint.  This is also true in Panphagia (uncoded) and Pampadromaeus (miscoded).  This would leave it as a phytodinosaur character, except that Turfanosuchus (miscoded) has it has well, and among 'paraornithischians' only Effigia has it (Silesaurus, Shuvosaurus and Lotosaurus miscoded).  Daemonosaurus is correctly coded as having it.
37. Orbit enters anterior half of skull.  Within 'paraornithischians' this is only true in Lotosaurus, not Silesaurus (unknown, miscoded), Shuvosaurus (unknown, miscoded) or Effigia (miscoded).  Otherwise it would fit combining derived sauropodomorphs with ornithischians.  Though note Daemonosaurus lacks it and is again miscoded.
95. Interpterygoid vacuity tapers sharply anteriorly.  This was coded as absent in Turfanosuchus (present), Gracilisuchus and Herrerasaurus (unknown).   In reality, all examined taxa except Gracilisuchus have it or could have it.
113. Caniniform teeth absent in maxilla.  This is here mostly due to misunderstandings on Peters' part.  He codes the toothless scorable poposaurs as having it, but as they lack teeth of any sort, they should be coded inapplicable for whether those teeth are caniniform.  More importantly, the character is used by authors Peters' cites for it such as deBraga and Rieppel (1997- ch. 95) in the sense that a taxon has caniform teeth when their size is greatest near the middle of the maxillary tooth tow (as opposed to anteriorly, or all being equally large).  So we can code taxa for this character even when they lack elongate pointed teeth that are usually thought of as caniniform.  When we do so, we see that Silesaurus and Massospondylus kaalae truly do lack caniniform teeth, but that Pantydraco doesn't preserve the anterior maxilla and that ornithischians have teeth near the middle largest so actually have a caniniform region.  Expanding to taxa not included by Peters, Sacisaurus still has a caniniform region, as does Efraasia, while Plateosaurus is debatable (first tooth slightly smaller).  So it seems that Silesaurus and plateosaurs evolved this convergently.  Peters correctly codes Daemonosaurus has having caniniform teeth.
115. Lateral teeth blunt.  This is part of a composite character involving sharpness, serration size and crown base constriction.  It's a good example of why such characters are bad, because while Scelidosaurus is coded as blunt (state 1), its teeth are also multicusped with a constricted base (state 3).  In any case, we'll look past the technical errors here and check the distribution of teeth with large serrations and constricted bases, which is what makes the teeth in question blunt compared to a typical archosauriform tooth.
For serration size, Panphagia and Pampadromeus are miscoded as being different than e.g. Thecodontosaurus and Pantydraco.  Yet Peters leaves Saturnalia uncoded, when it actually has small serrations.  Another thing Peters leaves out is that the fangs of Heterodontosaurus have tiny serrations like carnivorous archosaur teeth, so it has both states.  He correctly codes Silesaurus and Daemonosaurus as lacking it, but leaves Poposaurus uncoded when it lacked it too.  All of this combines so that it is equally parsimonious being a phytodinosaur character or developing convergently in Pampadromaeus+Panphagia, sauropodomorphs, ornithischians and Pisanosaurus in his tree.
The situation is similar for constricted crown bases- Pampadromeus and Panphagia were miscoded as lacking them, Heterodontosaurus actually has both states due to its fangs, Daemonosaurus was correctly coded as lacking it, and Poposaurus lacks it too but was left uncoded.  Though because Saturnalia and Silesaurus both have this character, it's slightly more parsimonious as a phytodinosaur synapomorphy reversed in a few cases.
What complicates the situation are the silesaurids Peters didn't include.  Sacisaurus and Diodorus both have large serrations, but Asilisaurus has small ones like Silesaurus itself.  Sacisaurus and Diodorus also have constricted roots, but Asilisaurus is unique among silesaurids (ignoring Lewisuchus of course) in lacking them.  This leaves both characters changing 6.5 and 4.5 times respectively if they are phytodinosaurian synapomorphies that reversed in some members.  They're much simpler in the standard phylogeny, where they each evolve three times (in derived silesaurids, derived sauropodomorphs and derived ornithischians), and the basal members of each (Asilisaurus, Silesaurus, Saturnalia, heterodontosaurids) show intermediate conditions.
122. Dentary extends into coronoid process.  Back to simple problems.  Peters miscoded Pantydraco (unpreserved), Daemonosaurus, Lotosaurus and Effigia as having this.  He also didn't code Silesaurus and Shuvosaurus for it, but they lack it too.  It's actually only present in ornithischians and some anchisaurs like Massospondylus, but not more basal sauropodomorphs unexamined by Peters like Efraasia and Plateosaurus, or even some taxa closely related to Massospondylus like Adeopapposaurus and Mussaurus.
128. Mandible straight ventrally.  Another complicated character because Peters' divides this into several states, with the anterior and posterior mandible being straight, concave or convex.  It's full of homoplasy in Peters' codings, with Silesaurus+Lotosaurus and ornithischians reversing to the 'straight anterior convex posterior' state he codes Panphagia and Pampadromeus for.  But as with so many of these characters, Peters miscodes a large number of taxa.  Saturnalia, Pantydraco and Massospondylus kaalae actually have concave anterior mandibles, his basalmost ornithischian Scelidosaurus and Pisanosaurus have convex ones, and Lotosaurus' and Shuvosaurus' are concave.  Posteriorly, none of his sauropodomorphs can be coded, Silesaurus' is convex, Lotosaurus and Effigia have both straight and convex margins and his basalmost ornithischians have both straight and convex (Scelidosaurus), concave (Heterodontosaurus) or convex (Agilisaurus) margins.  There's a huge amount of homoplasy and numerous taxa vary simply between individuals or sides of the head.
131. Cervicals decrease in size anteriorly.  Besides the atlas and often the axis, this isn't true in any examined taxon.  Peters simply miscoded them.
137. Mid-cervical centra (here c5-7 are used) shorter than mid dorsal centra (here d7-9 used).  This was miscoded by Peters as true in sauropodomorphs (ironically famous for their long necks).  It's found in Poposaurus, Lotosaurus and ornithischians (for at least some vertebrae), and is more parsimonious even in his tree as being convergent in these groups than being a phytodinosaur character lost in sauropodomorphs, silesaurids and shuvosaurs.
142. Lumbar area ("shortened to missing dorsal ribs") absent.  This is a hard one to code due to preservation and definition.  Very few taxa preserve the posteriormost dorsal ribs, and they're shorter than others as a rule.  It might be easiest just to note that Peters' coded many taxa whose posteriormost ribs are unpreserved and/or present in reconstructions, which accounts for all six non-ornithischians coded as lacking a lumbar area by Peters.  Also Scelidosaurus, Peters' basalmost ornithischian, was described as having a lumbar vertebra by Owen, while Plateosaurus has one or two dorsals without ribs as well.
144. Three to four sacral vertebrae, which if ordered would be "three or more sacral vertebrae".  Herrerasaurus was miscoded as having less, while Pseudolagosuchus was miscoded as unknown, but has two.  This means it now takes 4 steps in Peters' tree, while it would take five for Dinosauria to start with three sacrals and reverse in Marasuchus, Tawa and silesaurid Pseudolagosuchus.  But also note basal silesaurid Asilisaurus only has two, while theropod Eodromaeus has three, as does Eoraptor which wherever it goes seems close to theropods or Panphagia and Pampadromeus.  This leaves it more likely three sacrals are primitive for dinosaurs in Peters' tree, and more likely Silesaurus got its large sacral count convergently in reality.
175. Metacarpal II longest.  This is actually a fine character for grouping Sauropodomorpha and Ornithischia (note the condition in Pampadromaeus and Panphagia aren't known, so it doesn't support them being basal to both; indeed, the similar but unincluded Guaibasaurus also has it).  Of Peters' coded 'paraornithischians' though, Poposaurus has it but Lotosaurus is miscoded as having it (based on a sculpted museum mount I assume).  So strict Phytodinosauria sure, but that plus poposaurs is ambiguous.
195. Tibia shorter than femur.  This also works in Peters' tree, helped by his placement of Scelidosaurus basally in Ornithischia, as others' have long tibiae.   But it takes the same number of steps in the standard tree, where dinosauriforms have long tibiae, and it reverses in Silesaurus, Herrerasaurus, sauropodomorphs and Scelidosaurus (+ eurypods).
199. Tibia less than twice ilial length.  Hey, didn't we just look at tibial length? *cough correlated cough*  But that doesn't even matter here, because this is also true in Pampadromaeus (left uncoded by Peters), Herrerasaurus (miscoded) and possibly Tawa (the schematic reconstruction suggests so, but I would leave it uncoded for now).  So of included taxa, it's only not true in Marasuchus, Silesaurus and Panphagia.  Invalid.
Results- Of fifteen characters, Daemonosaurus has one (nasals widest in middle; which would work for Sauropodomorpha+Ornithischia) and a half (anterior mandible straight ventrally), though the latter doesn't actually support any similar clade.  It lacks five.  Nasals widest at midpoint would support Sauropodomorpha+Ornithischia; orbit enters anterior half of skull would support derived sauropodomorphs plus ornithischians; metacarpal II longest would support either variant; premaxilla a third or more of preorbital length and (somwhat ambiguously) short mid cervical centra would support Poposauridae+Ornithischia.  Large serrations and constricted crown bases would indeed resolve at the previous node (Phytodinosauria) in Peters' tree, but requires less homoplasy in the standard tree.  A short tibia also resolves here, but is equally parsimonious in the standard phylogeny.

'Paraornithischia' (Pisanosaurus, Poposauridae, Silesauridae)
28. Maxilla ventrally straight.  Miscoded in Pisanosaurus (unknown), Silesaurus, Shuvosaurus (unknown) and Effigia, so only Lotosaurus is correctly coded.  Varies so much that either a straight or convex maxilla might be basal for Saurischia, Dinosauria, Silesauridae+Dinosauria or Dinosauriformes.
112. Maxillary teeth shorter than twice their FABL.  Note Peters wrongly codes the toothless poposaurs as having this when they should be inapplicable.  More generally, this true for some teeth in basically every taxon known from a good sample size, including standard carnivorous taxa like theropods.  It's only true for all maxillary teeth in some derived ornithischians.
117. Laterally placed surangular ridge in dorsal view.  This was an example of characters Peters misunderstood and thus coded inaccurately (see part 1 of my critique).  In actuality, all examined taxa have this ridge.
119. Dentary straight.  Pisanosaurus miscoded.  While true in poposaurs and silesaurids, this is also primitive for dinosaurs in general and found in Arizonasaurus.  Peters didn't find this due to miscodings (e.g. Scelidosaurus, Heterodontosaurus, Panphagia) and missing taxa.
167. Olecranon process absent.  Miscoded in Pisanosaurus (unknown), Lotosaurus (unreported), Poposaurus (unreported).  It's only true in Silesaurus, Effigia, Gracilisuchus and some ornithischians (Scutellosaurus and Agilisaurus+Hexinlusaurus).  Thus it fits with the silesaurid+poposaur clade in Peters tree, though in the standard tree the olecranon in basal rauisuchians sensu lato and crocodylomorphs indicate convergence.
178. Four phalanges in manual digit IV (up from three or less).  Miscoded in Pisanosaurus (unknown), Lotosaurus (five phalanges) and Effigia (unknown).  Correctly coded as absent in Poposaurus (three phalanges), and no other 'paraornithischians' can be coded.  This leaves the state untrue for any taxon.
205. (Preseumedly large) calcaneal tuber proximally oriented.  Miscoded in Pisanosaurus (absent) and correctly coded as absent in silesaurids.  Thus it's only present in poposaurs, which in standard trees is because they're outside Avemetatarsalia and doesn't require a reversal then.
Results- Only the absent olecranon would work for this clade in Peters' tree, which is only known in Silesaurus and Effigia.  Note Pisanosaurus was miscoded for all except one which is present in all examined taxa (and coded correctly by accident by Peters as he was looking at the wrong structure), and another which actually varies in almost all taxa (and may vary in Pisanosaurus if anterior maxillary teeth were preserved, as these are the elongate ones in basal ornithischians).  Thus there's no evidence Pisanosaurus is closer to poposaurs and silesaurs than to ornithischians.

(Poposauridae+Silesauridae)
84. Postorbital extends posteriorly less than halfway through the parietal.  This was only coded as true in Lotosaurus and Effigia and is miscoded in both anyway.  It's actually not found in any included taxon except Terrestrisuchus.
115. Maxilla toothless.  Poposaurus was miscoded as unknown when it has maxillary teeth.  Being absent in silesaurids, it's only present in Lotosaurus and shuvosaurs, which I have no issue with and takes less steps in the standard phylogeny than Peters'.
118. Paired predentary.  A fictional element found in no animal I know of. Peters just imagines predentaries are fused to dentaries in these taxa.
135. Cervical ribs of average width and at high angle to neck.  Both width and orientation miscoded in Lotosaurus, Silesaurus and Effigia (unknown).  Only Gracilisuchus, some Lotosaurus ribs and some Hexinlusaurus ribs are not slender, and no examined taxon has high-angled ribs.
140. Cervicals 6-9 without offset central faces.  Miscoded in Silesaurus, and I have no issue with poposaurs having it, since it's generally considered the primitive condition modified in dinosauriforms.
160. Clavicle present (and shorter than scapula, as if that needed to be said).  This is only here because Peters miscoded a ton of taxa as lacking clavicles when they're just incompletely preserved (also note no silesaurid can be coded).  In actuality, only some crocodylomorphs and ornithischians have good evidence for missing clavicles, while saurischians retain them.
181. Manual unguals blunt.  Only known in poposaurs (though Effigia was miscoded).
Results- Not a single character is even present in silesaurids.  Perhaps the biggest fail of any Peters clade examined here.

(Lotosaurus+Silesauridae)
14. Premaxilla-maxilla notch over 45 degrees.  This entire character is miscoded by Peters.  In actuality, all examined taxa lack a notch except Tawa+Coelophysis, Pampadromaeus and heterodontosaurids (with Eoraptor and Daemonosaurus being intermediate).
15. Premaxilla ventral surface angled anteroventrally.  This is untrue in silesaurids and, contra Peters, present in Effigia and probably Poposaurus.
31. Orbit shorter than postorbital skull length.  This is unknown in silesaurids.
44. Frontal not wider posteriorly than anteriorly.  This is widely miscoded by Peters and actually only present in Gracilisuchus among examined taxa.
125. Posterior mandible of even depth anteriorly and posteriorly.  Intrue, both Silesaurus and Lotosaurus have mandibles deepest in the middle, which is true for most taxa.
128. Ventral mandible edge straight anteriorly and convex posteriorly (as opposed to completely straight).  Lotosaurus actually has a concave anterior mandible, with a (sometimes) convex posterior mandible is typical of all but some herbivorous dinosaurs and is also found in some specimens of the only other scorable poposaur, Effigia.
145. Sacral neural spines taller than acetabulum.  True in Lotosaurus and Silesaurus, but unknown in Pseudolagosuchus (unknown), and absent in the unexamined silesaurid Asilisaurus.  This makes the character ambiguous. 
149. Gastralia absent.  Peters miscoded Silesaurus (absent), Pseudolagosuchus (unknown) and Lotosaurus (unknown).  Another instance of Peters assuming partially preserved skeletons or mounted specimens include all originally present elements.
170. Forelimb between 55 and 100 percent of hindlimb length (instead of being shorter).   The ratio in Lotosaurus is unreported and was guessed at by Peters from mounts of uncertain completeness.
183. Preacetabular process truncated.  The condition in Lotosaurus is unreported and was guessed at by Peters from mounts of uncertain completeness.
200. Iliofibularis tubercle a 'spot'.   Silesaurus, Pseudolagosuchus and Lotosaurus all actually have a crest-like tubercle, so are miscoded.
210. Metatarsus less than half of tibial length.  Miscoded in Silesaurus, Poposaurus and Effigia, so only untrue in Effigia among these taxa. 
Results- Only the long sacral neural spines even ambiguously link Silesaurus and Lotosaurus.  The rest of the characters are highly miscoded.

(Sauropodomorpha, Daemonosaurus, Ornithischia)
33. Naris larger than antorbital fenestra.  This was miscoded for Daemonosaurus, Saturnalia (unknown) and Pantydraco.  It's actually only present in ornithischians and some plateosaurs, as even taxa like the unincluded Riojasaurus, Sellosaurus, Lufengosaurus and Jingshanosaurus lack it.
41. Frontoparietal suture straight and longer than frontonasal suture.  Another composite character, the length aspect of which is true for basically all examined taxa.  Straightness does indeed work as a character for this clade though.
69. Squamosal and quadratojugal do not contact.  This is only here because Peters miscoded theropods as having contact and sauropodomorphs, Heterodontosaurus and Lesothosaurus as lacking contact.  It's actually only true in theropods and some ornithischians (Scelidosaurus, Agilisaurus+Hexinlusaurus).  Miscoded in Daemonosaurus as it is actually not known.
127. Retroarticular process angles ventrally.  Widely miscoded and only actually present in Hexinlusaurus among included taxa.
179. Manual digit IV reduced in width.  This is true in almost all dinosaurs and was actually miscoded by Peters as absent in Pisanosaurus (actually unknown), Poposaurus and Effigia.  So in his tree it would actually diagnose Dinosauria and reverse in Lotosaurus and (miscoded) Lesothosaurus.  Note it is equally parsimonious for it to converge in derived poposaurids and dinosaurs as in the standard phylogeny.
191.  Acetabulum at least partly perforated.  Similar to the previous character, this is true in all dinosaurs and was miscoded by Peters as being absent in the examined poposaurs.  It was also miscoded as lacking in Pampadromaeus and Panphagia, and may even be present in Silesaurus (though lacking in the unincluded Asilisaurus and probably Sacisaurus). 
216. Phalanges of pedal digit IV subequal to metatarsal IV.  Here's an example of how not ordering your characters can lead to problems.  In Peters' data, the outgroup and 'paraornithischians' have long phalanges, sauropodomorphs and ornithischians have subequal phalanges and theropods have short phalanges.  PAUP doesn't realize which of these are most similar, but if it were ordered PAUP would know this means theropods are most like sauropodomorphs and ornithischians.  In any case, Peters miscoded Herrerasaurus which has subequal phalanges too, and Saturnalia, Lesothosaurus and Hexinlusaurus (left uncoded) which all have shorter phalanges.  The latter examples, along with Pisanosaurus and unexamined Tianyulong having short phalanges, means this is probably the basal state for ornithischians and sauropodomorphs.  Yes Peters has Scelidosaurus basal in Ornithischia, but also miscoded it as subequal when it actually has long phalanges, so doesn't fit with subequal being basal for sauropodomorphs+ornithischians anyway.  As theropods are subequal (Herrerasaurus, Dilophosaurus) or short (Coelophysis, Liliensternus, Procompsognathus), this leaves it ambiguous whether short phalanges are a phytodinosaurian character.
Results- Daemonosaurus has one of these (straight frontoparietal suture), which is indeed valid for this clade.  It lacks two, neither of which really diagnoses the clade anyway.  Besides the straight frontoparietal suture, only the short pedal digit IV phalanges even ambiguously works for this clade.

(Daemonosaurus, Ornithischia)
114. Last maxillary tooth placed at mid orbit (instead of anterior orbit).  This works, though note it's also present in Eoraptor, wherever that taxon goes (Peters seems to think Theropoda).
Results- Yes, this single character is valid for this clade.


Total Results- There are several heterodox aspects that should be evaluated.

First is that 'paraornithischians' are closer to phytodinosaurs than to theropods.  This is supported by external naris being angled more dorsally, which takes one more step in the standard phylogeny.  A short tibia takes equal steps, and while large serrations and constricted crown bases work for this node in Peters' tree, they are better explained by the standard tree.  Even if we just correct Peters' characters, this grouping takes 12 more steps in his trees.  When I added new characters and taxa, it took 34 more steps.  In Nesbitt's analysis with far more anatomical data (even excluding taxa Peters doesn't see as avemetatarsalians, like pterosaurs and lagerpetonids), it takes an astounding 76 more steps.  Utterly rejected.

Regarding Paraornithischia itself, no characters support placing Pisanosaurus in it, and this is 13 (Peters' characters corrected), 26 (Peters' corrected matrix with more data added) and 50 (Nesbitt's reduced matrix) steps longer than the most parsimonious tree.  Pisanosaurus being a paraornithischian is very strongly rejected.

For a more reduced Paraornithischia containing only silesaurids and poposaurs, only two characters could  plausibly support this in Peters' matrix.  Silesaurus could share an absent olecranon with poposaurs (only known in Effigia), then ambigiously share long sacral neural spines with Lotosaurus (unknown in Poposaurus and Arizonosaurus, though probable in both considering other neural spines; absent in shuvosaurs).  This is incredibly weak support compared to the numerous characters placing Silesaurus close to dinosaurs and poposaurs in Pseudosuchia.  Enforcing poposaurs and silesaurids as a monophyletic group takes 8 more steps in Peters' matrix both with corrected codings and once the extra characters and taxa are added (in both cases they are sister to Dinosauria and Pisanosaurus is ornithischian), and 41 more steps in Nesbitt's reduced matrix (where they are sister to Dinosauriformes and Pisanosaurus again is ornithischian).  So strongly rejected again, though at least leaving Pisanosaurus out made it a tad less homoplasious.

Peters places Lotosaurus closer to silesaurids than to poposaurs.  Given the liklihood Poposaurus has tall sacral neural spines noted above, even this one remaining possibly applicable character is questionable.  This grouping takes 15 more steps using Peters' matrix with corrected codings, where the Lotosaurus+silesaurid clade ends up in Crocodylomorpha.  It takes 19 more steps once the extra characters and taxa are added, where the clade falls out between Crocodylotarsi and poposaurs on the grade leading to dinosauriforms.  No other silesaurids follow.  Finally, it takes 45 more steps in Nesbitt's reduced matrix, where it's apparently easiest just to let Lotosaurus be placed as the basalmost silesaurid.  Another strongly rejected concept, which when taken together with the previous two paragraphs shows all parts of 'Paraornithischia' besides Silesauridae and Poposauridae each being real is unlikely in the extreme.

Back to actual dinosaurs, there is Peters' idea derived sauropodomorphs are closer to ornithischians than to Panphagia and Pampadromeus.  This is supported by orbit enters anterior half of skull, a straight frontoparietal suture, (ambiguously) posterior mandible deepest anteriorly and (ambiguously) tibia shorter than femur, which are outweighed by the sauropodomorph characters not included by Peters.  Just fixing Peters' characters leaves this 5 steps less likely than sauropodomorph monophyly, and adding taxa and characters brings it to 11 steps less likely.  But if you try this in a dedicated sauropodomorph matrix like Cabreira et al. (2011), it's a whopping 25 steps less parsimonious.  Very unlikely to be true.

Then there's Sauropodomorpha+Ornithischia itself.  This has actually been suggested by other workers before, unlike most of Peters' unusual ideas (technically, Peters doesn't support the standard version since Pisanosaurus, Pampadromaeus and Panphagia are excluded, and Daemonosaurus is included).  This is supported by nasals widest at midlength, (ambiguously due to Saturnalia) enlarged serrations, teeth constricted at base, metacarpal II longest and (ambiguously) short phalanges of pedal digit IV.  But these are outweighed even in his own matrix, where correcting codings makes it 2 steps less parsimonious.  Adding characters and taxa brings this up to 8 steps.  In Nesbitt's matrix (again only using taxa Peters did in this part of his tree), it takes 13 more steps.  So this non-Petersian Phytodinosauria isn't completely rejected, but does seem quite unlikely.

Finally, there's the idea Daemonosaurus is a member of any of the above clades instead of being a theropod.  Of all of the above valid characters, it only has nasals widest at midpoint (sauropodomorphs+ornithischians), straight frontoparietal suture (derived sauropodomorphs+ornithischians) and last maxillary tooth placed at mid orbit (ornithischians).  But it lacks large serrations, constricted tooth crown bases, orbit enters anterior half of skull, and posterior mandible deepest anteriorly, and the above clades are probably not real anyway.  We saw before forcing Daemonosaurus to be ornithischian took 7-10 more steps depending on the matrix used.  Forcing it to be phytodinosaurian takes 2 (sister to remaining phytodinosaurs), 10 (it's sister to Eodromaeus with both sister to other phytodinosaurs, and theropods are paraphyletic to the latter) or 20 (herrerasaurids become basalmost phytodinosaurs, followed by Daemonosaurus+Eoraptor) more steps in Peters' corrected, Peters' corrected with added data, and Sues et al.'s reduced analyses respectively.  So far from helping Phytodinosauria, placing Daemonosaurus in that clade makes it 0, 2 or 4 steps less likely to exist.  The most you can say for Peters' idea is that in the quite unlikely chance Phytodinosauria is real, making Daemonosaurus the basalmost member is only slightly more problematic, though taxa he considers theropods such as herrerasaurids, Eoraptor and/or Eodromaeus would follow.

As Daemonosaurus is so emphasized by Peters, which characters suggest it is a theropod instead of an ornithischian?  We previously saw it took ten more steps to place Daemonosaurus in Ornithischia in the full matrix used here, and seven more in Sues et al.'s (2011) original matrix.  Some of the characters suggested by Sues et al. have more homoplasy when examined in the larger matrix, but several have relatively clear distributions.  Remember despite the homoplasy noted below, these were still found to diagnose the clades when all of Peters' characters were included too.  Characters that could also work for an ornithischian placement aren't noted.

Saurischia
- Ventral ramus of the opisthotic not covered by the lateralmost edge of exoccipital in posterior view. Reverses in Massospondylus kaalae and avepods.  Not used by Peters.

Saurischia/Theropoda
- Maxilla ventrally convex.  Exactly which node this belongs to is uncertain as Eoraptor and Pampadromaeus have it, but other sauropodomorphs don't.  Tawa is miscoded by Peters.

Theropoda
- Snout roughly triangular. As in Tawa and Coelophysis, which are more derived than Eoraptor, Herrerasaurus and Eodromaeus in this tree.  Miscoded in Daemonosaurus by Peters.
- Sharp jugal ridge. In all examined theropods including Eoraptor, also in some Scelidosaurus specimens.  Not used by Peters.
- Deep pneumatic fossae in cervical centra. Also in sauropods.  Not used by Peters.

Sister to Tawa
- Less than four premaxillary teeth.  Also in heterodontosaurids, but as they're not sister to Daemonosaurus or even basal in Ornithischia in Peters' trees, this doesn't matter.  Miscoded in Daemonosaurus and Tawa by Peters.
- Jugal, anterior extent of the slot for the quadratojugal: at or anterior to the posterior edge of the dorsal process of the jugal.  Also in Pampadromaeus, some Scelidosaurus and Massospondylus carinatus specimens.  Not used by Peters.
- Exoccipital, lateral surface: without subvertical crest (metotic strut). Unique among dinosauriforms.  Not used by Peters.

"You have your assignment: Nest Daemonosaurus with theropods while including Heterodontosaurus and Massospondylus. That’s a half-dozen to a dozen taxa at most to deal with. Then we’ll compare answers."

Assignment complete, Peters.  Comparison indicates massive miscoding by you, a lack of included taxa and characters, and in this particular case using a largely fictional skull for Daemonosaurus.  Almost like what I figured out by just looking at a portion of your dataset without reanalyzing the entire thing...


And that wraps things up as far as Peters' Dinosauria goes.  Unfortunately, despite my utter demolishing of his analysis, he's unlikely to fix it.  So many of the problems are due to a third of the characters being miscoded, but while he seems willing to fix a few, for most he remains stubbornly convinced that his tracing is better than what's indicated by the actual experts who saw and studied the material.  Sure you can tell Peters where he gets things wrong, but he'll consider that worthless.  Sometimes he requests you take the time to trace your interpretation for him, but his errors are so numerous and his understanding of anatomy so poor that it would be a full time job for anyone to educate him.  He also remains intransigent regarding how he should fix his character formulation and ordering.  Just doesn't care.  Similarly, Peters is blase about his potential for miscoding by using reconstructions and sculpted mounts.  It will balance out he says, because his terribly miscoded analysis has falsely assured him that most nodes are strongly supported, making noise unimportant.  He's unconcerned with adding more characters because he just wants to see what his 228 will find, apparently not realizing limiting it to such a small amount makes the result almost worthless.  It's like my coding new taxa into small older analyses on the Database- it's fun and interesting to see how taxa fall out using classic data, but I would never claim e.g. segnosaurs are closer to caenagnathids than to oviraptorids just because that's where they ended up when added to Gauthier's 1986 matrix.  Even if his whole topology is shown to be based on erroneous data, he'll get things backwards and compare the resulting trees to see how subjectively similar related taxa seem to him.  As if the result matters when the data used to get it are wrong.  So what point is there in trying to educate someone who refuses to learn?  While I was naive in thinking a detailed critique just might show Peters how much more work he needs to do before his analysis is useful, I hope this will at least help those amateurs who think Peters' ideas have merit and fall for his "I'm a rebel underdog whose ideas are disliked by the experts because they're blinded by tradition and conformity" schtick.  All of the information in this series of posts (and more) will be uploaded to the Database in the Evaluating Phylogenetic Analyses section.  Until Peters shows willingness to do more than correct a few scores, it's just a time sink to engage him in detail.  Hope you all enjoyed the series.

Monday, April 15, 2013

A conversation with Larry Martin

Jason Brougham has a post at his blog discussing Larry Martin, one of the BAND* leaders and later MANIAC** who passed away in March.  I met Larry at SVP 2002, where my reaction was-

"I had a good discussion with Martin. He's quite a nice guy, it's a shame he's so wrong about almost everything dinobird-related. Seems he officially believes that maniraptorans are birds, but that other coelurosaurs are unrelated. He terms this the "Paulian hypothesis", though it's only similar in thinking of maniraptorans as 2F and preferring the arboreal origin. In his mind, maniraptorans and other birds evolved from a Longisquama-like creature. And like Czerkas believes, some taxa (eg. Scansoriopteryx) are just too hard to place as birds or dinosaurs until a more in depth study is completed. The same old ABSRD arguments were brought up (digit homology, terrestrial theropods, Yixian feathers being collagen, etc.). He referred to both ichthyosaurs and Psittacosaurus as having structures like those found in Sinosauropteryx et al., which is just plain wrong of course. In any case, he does have an interesting eumaniraptoran that will be out soon."

That eumaniraptoran was later named Microraptor gui, btw.  As I recall, he said the leg remiges proved M. gui wasn't terrestrial, therefore birds evolved flight from trees, and therefore birds aren't dinosaurs.  I then said I have no issue with some dinosaurs being arboreal, which I think he was surprised by.
* Birds Are Not Dinosaurs, contrasted with BAD being Birds Are Dinosaurs
** Maniraptorans Are Not In Actuality Coelurosaurs

I feel Jason's post bends over backwards to seem fair, but also that it leaves out Martin's greatest positive influences on paleontology. 

He first writes about Martin being on "our side", which seems to be opposed to creationists, since "He spent his life studying evolution, observing fossils, and we all agreed that dinosaurs and birds were closely related archosaurs."  It seems trivially true that anyone will be on your side about some value or opinion.  Indeed, even evangelical creationists are on "our side" for wishing to teach people the truth, we just disagree on what that truth is.  In any case, as a scientist I think he would have been proud to be against "our side" when it came to heterodox ideas he strongly defended.

Jason believes Martin had a beneficial effect on bird origins research in that BADists "wrote some papers they would not have written unless they felt the need to answer Martin. In the process, progress was made on quantifying the ‘temporal paradox’ issue, and everyone’s hypotheses became more precise and explicit."  But I honestly can't think of many examples where Martin's criticisms directly led to science examining the issue more closely to result in more discoveries, they were mostly just pointed out to be based on fallaceous reasoning or inaccurate interpretation.  If we look at his classic (1983) arguments for BAND, the temporal gap problem had been answered perfectly years earlier by Ostrom (1976).  I don't see Brochu and Norell's (2001) quantification paper as doing more than telling us what we already knew, i.e. the gap between the Tithonian and Aptian is less than between the Tithonian and Triassic.  The arboreal vs. terrestrial rationale actually hindered our side in my estimation, as the BAD response was to double down on the terrestriality of Archaeopteryx and flight origins instead of accepting arboreal theropods could exist, which only occured decades later when scansoriopterygids and microraptorians were found.  These examples can be contrasted with e.g. Jones et al. (2000) instigating research on Longisquama's parafeathers or Burke and Feduccia (1997) instigating research on digital development.

The next argument is that some recent large morphological analyses (e.g. Livezey and Zusi, 2007) have shown Martin was right that convergent evolution could confound cladistic analyses.  I don't think anyone's ever doubted this in principle, as analyses do differ so some must be wrong.  Moreover, the ways such analyses were shown to be confounded were never used by Martin (more characters and taxa, molecular analyses, biogeography, etc.), with his naive falsification and evolutionary scenario alternatives ignored (Makovicky and Dyke, 2001).  Sure Martin (1997) was right(er) than Chiappe that alvarezsaurids are ornithomimosaurs instead of avialans, but it was for all the wrong reasons, so should that really count scientifically?

Finally, Jason says of Martin's 2005 transition to MANIAC that "I have heard few mention that, in this 2004 paper, Martin rather quietly gave up on that view [that dromaeosaurs were only distantly related to birds]."  "... this demonstrates that Martin was open-minded enough to be persuaded by the evidence and to reverse himself on a major part of his hypothesis."  While I do give credit to Martin for finally recognizing this relationship, the details make it look less noble.  First note Martin never actually admitted his quarter century of arguments against Archaeopteryx and Deinonychus being similar were fallaceous, or that this calls into question his entire methodology.  Indeed, he still uses most of them in that very paper, seemingly without realizing his new phylogeny cancels them out (e.g. if maniraptorans are all considered birds, then deriving this whole group from something standard like Ornitholestes eliminates any temporal paradox).  Second, he (as with Feduccia) framed neoflightless maniraptorans as a surprising compromise, when by that time most theropod workers considered it possible if less parsimonious.  The feeling from MANIAC papers is that taxa like Microraptor and Scansoriopteryx have sent everyone back to the drawing board, when BADists have actually been relatively unaffected besides the character distributions in our cladograms.  The impression is an attempt to save face while admitting as little error as possible, which seems distinctly unscientific to me.  Still, I'll say Martin's work was less polemical than Olson's or Feduccia's, the latter's 2013 paper being more unhinged than anything Martin wrote.

Well, that's a lot of criticism of a dead man, but I've never been one to follow the tradition of emphasizing the positive aspects of the deceased while downplaying the negatives.  Yet positives did exist, and I think his greatest legacies in science have yet to be articulated.  Foremost, Larry worked on a LOT besides bird origins.  While I'm not qualified to review e.g. his saber-tooth cat work, his hesperornithine work such as the 1976 osteology of Baptornis was quite good.  He continued such studies until the end of his life, such as his 2011 ornithuromorph predentary paper with Zhou that analysed an often ignored element.  Which brings us to Martin's other great legacy- his students.  I know at least Zhou and Witmer trained under Martin.  Zhou's been instrumental in expanding our knowledge of Jehol bird diversity.  As for Witmer, he has this touching summary on his institutional biography page, which shows my two favorite aspects of his excellent studies are due to Martin-

"I also owe a huge debt to Larry Martin, who got me interested in so many things.  He always encouraged me to look at modern animals along side extinct groups, and this has emerged as my major research paradigm, these nascent ideas later evolving into the extant phylogenetic bracket approach.  And of course it was Larry who introduced me to the previously under-appreciated anatomical system of cephalic pneumaticity, which became my major research focus for a decade and a half.  Although Larry and I are often now pitted as opponents in the fierce debate on avian origins, he has been one of the most important positive influences in my career."

So there are my criticisms and accolades.  But there's one more story I'd like to tell.  Back in October 2004, I decided to write Larry regarding bird origins.  It wasn't meant as a harsh debate, and I wasn't so naive as to think I could do what two decades of published professionals couldn't, but I was intrigued to see where he was coming from and how he would respond to questions and facts that aren't brought up in his papers.  Note this was before his 2005 paper declaring his MANIAC status, but I knew he thought that way due to SVP 2002.  I reproduce the content of this email exchange below, with changes in color and font between Mickey Mortimer and Larry Martin.  I've left out most of the pleasantries preceding and following each actual email, and have formatted them so that e.g. when Martin has four paragraphs, my four paragraphs are responding to them in that order.  While publishing email exchanges without explicit permission is of course frowned on in most instances, I'm hoping my own feelings this is ethical once one party is dead are shared by the community.  I think of Martin's words being like the personal letters of historical scientists that are now public, and it's all scientific exchange matching what he published as opposed to personal details.  So enjoy.  I don't think I've ever read a conversation between a BADist and BANDit, which makes this rather unique.  For all the criticisms I piled on Larry, I don't know many other scientists who would take the time to write such a long and detailed exchange with an unknown undergraduate amateur.  He'll be missed.



I'm an undergraduate at the University of Washington
currently investigating theropod and Mesozoic bird phylogeny.  I heard you
gave a talk the other day where you advanced the hypothesis that
maniraptorans are in actuality birds, and that they are unrelated to
theropods.  I wonder if you could answer a few of my questions.
1. What characters show oviraptorosaurs and dromaeosaurs are birds, while
leaving ornithomimosaurs and tyrannosaurs in Theropoda?
2. What is the sister group of birds + Longisquama in your hypothesis?
3. Do you believe homeotic shifts occur in nature?
Thank you for your time.
 
I suppose that we could define birds as animals who have or
give evidence of have having an avian wing (primary feathers attached to the
middle digit with the palm extending between the middle and outer digits and
a digit reduction pattern of 2-3-4) all animals with such a wing may
reasonably be supposed to have shared a common volant ancestor.  This brings
dromaeosaurs and oviraptosaurs into the avian fold.  The sister group of
birds and Longisquama (including Longisquama relatives) would probably be
dinosaurs plus many of the more conventional thecodonts.  I am not sure how
crocodilians fit in.  In any case the connection would be very basal within
archosaurs.  Please notice that if dromaeosaurs are secondarily terrestrial
we can expect many reversals of polarity from earlier analysis.  The
frameshift hypothesis was badly crunched by developmental types in a recent
conferance and is only being maintained to preserve a theropod hypothesis
that now fatally flawed.
 
Accepting your definition of 'bird', why are tyrannosaurs and
ornithomimosaurs excluded?  We have no embryological evidence to know if
they possess digits I-II-III or II-III-IV, and no preserved manual
integument that would tell us if they had primary feathers (no manual
integument has been reported for Pelecanimimus, Dilong, or tyrannosaurids,
though they each preserve integument elsewhere).
What characters do dinosaurs and these 'more conventional thecodonts' share
than Longisquama and birds lack?
Finally, do you mean the particular digital homeotic shift theorized for
theropods was questioned in this recent conference, or the reality of
homeotic shifts as a whole?  In either case, I'd be interested to know the
name of the conference, so that I can examine the references.

Thanks for the reply.  I agree dromaeosaurs and oviraptorosaurs were
secondarily flightless, which indeed leads to reversals of character
polarity from prior analyses. 

Tyrannosaurs are excluded because their hand as restored is not avian,
(they are not maniraptoran), maybe evidence could be provided that they
were once that way, but why worry until there is some reason.  They also
lack a lot of the other cranial features, etc.  Recent description of so
called protofeathers on a tyrannosaurid would seem to be clearly muscle fibers 
or ligaments and is probably better seen as evidence for a lack of feathers.
 
Ornithomimosaurs (including Mononykus) have derived hands with enlarged
MTCI and a reduced wrist.  I can't see how it relates to the avian hand
and think that they are best compared to compsognathids. I think in terms
of the general story, it makes very little difference where these taxa go. 

If Tyrannosaurs are carnosaurs, we know that at least some carnosaurs have
typical dinosaur rosette scale patterns and this may be a dinosaur
synapomorphy.  They also have interdental plates covered by a 
superdentary bone and this would make them very unlikely birds.  They also
have a long postacetabular ilium, a derived feature found in the earliest dinosaurs and absent from
all of the very early birds.
 
The conference was the Society of Avian Paleontology and Evolution. 
Essently none of the developmental types thought much of frameshift
although one tried to preserve bird/theropod by postulating a presently 
unknown ancestral six fingered hand.  You have to ask yourself about 
the credibility of an idea that requires a lot of ad hoc new science
to keep it alive. 
 
I accept your proposal tyrannosauroids lack some maniraptoran characters.
What prevents them from being maniraptoran relatives though?  Branching off
the bird line before maniraptorans developed all of their distinctive
traits?  They do share a lot of characters with maniraptorans, after all.
Regarding Dilong, what features do its supposed feathers have that would
make them less likely to be feathers than the identical structures in
Sinornithosaurus' holotype, or the plumulaceous body feathers of Caudipteryx
and Protarchaeopteryx (besides the fact you know these last three are
maniraptorans; that would be circular reasoning, after all)? 

Check out Ingenia for a maniraptoran example of enlarged mcI, though I admit
no maniraptoran reduces its carpals like ornithomimosaurs.  I agree with you
about Mononykus and other alvarezsaurids being closely related to
ornithomimosaurs.  I even wrote a post to the Dinosaur Mailing List in June
about how they were being put inside Maniraptora and Avialae due to flaws of
most cladistic analyses.  I also agree both groups are close to
compsognathids.  I think ornithomimosaurs+alvarezsaurids could be very
important for your hypothesis, since they show a number of characters you
think of as avian- no supradentary; serrationless teeth with basal
constriction; large sternum; ossified sternal ribs; costal facets on the
sternum; coracoid facets on the sternum; hollow keratinous integument.
 
Thanks for providing these specific characters.  Tyrannosauroids are almost
universally agreed to be coelurosaurs now, so we don't have to worry about
rosette scale patterns yet.  Velociraptor and Dromaeosaurus possess
supradentaries, so that can't bar tyrannosauroids from being on the bird
line.  If we use the ratio between postacetabular and preacetabular
processes to be an indicator of postacetabular length, the long
postacetabular processes of tyrannosauroids (~120-140%) are comparable to
some maniraptorans (Microvenator- 132%; Ingenia- 128%; Sinornithoides-
146%).  So that can't bar them from being on the bird line either.  Unless
you use a different ratio. 

I completely agree with the latter sentiment.  Unfortunately, the 2004 SAPE
abstracts aren't available yet, so I'll have to wait to see what new data
was presented.  It will be interesting to see how they deal with homeotic
shifts observed in human cervical number, which we can validate since we
know the geneology of the people in question.
 
please give citations for superdentaries in maniraptorans of any sort!
 
I think that the position of tyrannosaurs is quite controversal and it
doesn't make any difference to me how they fall out.
 
I am not convinced that any of the so called protofeathers are real.
Everybody has muscles and just because something is on a bird doesn't mean it's a feather. 
 
Gladly.  If you ever want confirmation of my statements, just ask.
Dromaeosaurus-
Currie, 1995. New information on the anatomy and relationships of
Dromaeosaurus albertensis (Dinosauria: Theropoda). Journal of Vertebrate
Paleontology. 15, 576-591.
Velociraptor-
Barsbold, 1983. Carnivorous dinosaurs from the Cretaceous of Mongolia. Joint
Soviet-Mongolian Paleontol. Expedition Trans. 19, 5-120. 

If tyrannosauroids are stem-birds, they would help us understand bird
evolution a lot.  There's a big gap between Longisquama and dromaeosaurs,
but tyrannosauroids help fill it out.  Besides, if theropods and
maniraptorans are so distantly related, it should be easy to tell them
apart, right?  Even in famous cases of convergence like thylacines vs.
canids, Notoryctes vs. talpids or hesperornithines vs. gaviiformes, the
differences are easy to see if one looks at skeletal details and not gross
resemblence.
 
This is a very interesting stance I've not seen in the literature.  How do
you reconcile this with the fact Shuvuuia's hollow cylindrical structures
are made of keratin (Schweitzer et al., 1999), and not collagen, actin or
myosin?
 
I doubt that any keratin has survived since the Cretaceous.  At least the
mollecular types don't buy into Mary's results.  Evidence that any of these 
structures were hollow is hard to come by.  All of the morphological structure 
is best expained by other interpretations. 
 
Quite a lot of the anatomy of dinosaurs has been uncertain, for instance
the two decade struggle to get the hand and wrist of Deinonychus correctly
interpreted. 

What are the citations for molecular biologists disputing Schweitzer et
al.'s paper, and/or keratin fragments being unable to survive since the
Cretaceous?  Schweitzer et al. state quite plainly "These fibers are seen to
be hollow, both grossly and in microscopic cross section," and photograph an
example in figure 2.  Which alternative explanation is better for clumps of
hollow fibers 200 micrometers in diameter which show smooth outer surfaces,
pithy inner surfaces, plate-like structures at one end, small filaments on
the outer surface, and consisting of microfilaments of 90-95 nanometers?
 
Agreed.  Even now, there are interesting new possibilities being raised,
like Gishlick's (2002) idea maniraptorans' distal carpal III is fused to the
base of metacarpal III.  As might be expected in e-mail discussions, it
seems a couple questions I'm curious about have got lost with all the
supradentary and keratin information.  Namely-
Why are some (which?) 'thecodonts' more closely related to dinosaurs than to
Longisquama+birds?
Given other examples of tetrapod convergence, shouldn't it be easy to tell
maniraptorans from theropods if they are so distantly related?

There has been quite a bit of rethinking of some of the earlier claims and
right at the mommentI think that real evidence of intact Tertiary proteins 
is pretty cutting edge.  Let's say for argumentation that she has keratin tubes, 
what feather structure could they be?  How would it form?
We should have some criteria for thinking something is a feather beyond
finding it on a dinosaur if we intend to show that dinosaurs have feathers.
 
I think that Roger Sawyer who provided the antigens is going to or has commented on the paper. 
Alan Feduccia probably could give you more commentary. 

Tyrannosaurs are so highly derived that I can't see that they would help much. 
Their wrists, hands, ankles. feet and skulls are all more derived and modified 
than the other tax we have discussed.
 
Well, these are only supposed to be keratin fragments, not intact keratin.
So that may make it more plausible to you.  If they are keratin tubes, the
feather structure represented would be the rachis and/or barbs.  Prum (1999)
goes into detail about how feathers form ontogenetically and perhaps
phylogenetically as well.
I agree with your statement we need non-circular reasoning if we wish to
show dinosaurs have feathers.  Prum has defined feathers in relation to the
feather follicle, making the first epidermal structure originating from a
follicle homologous to those in living birds the first feather.  I might
just be naive, but it seems to me any epidermal derivitive with a slender
cylindrical base necessitates a follicle to grow from.  So those of
stem-birds (like Confuciusornis and Sinornithosaurus), pterosaurs, and
Longisquama are are potential feathers, if we can establish homology with
birds' feathers.  This latter job is not easy, but is helped by phylogeny
and feather morphology.  As an example of the former, no non-feathered taxa
are closer to modern birds than Confuciusornis or Sinornithosaurus, so their
structures are phylogenetically congruent with homology.  An example of the
latter is the branching observed in the structures of Sinornithosaurus and
Dilong, which is unique to feathers.
 
Thanks.  I'll ask them. 

This doesn't appear to be true for basal taxa like Dilong.  Besides a few
characters (symmetrical premaxillary teeth in section; fused nasals; nasals
anteriorly convex in transverse section; pneumatic articular; preacetabular
notch on ilium), it looks like a good stem-bird.  Sure, it probably lacked a
true semilunate and other maniraptoran characters, but so did Longisquama.
 
If I understand Prum's model, his protofeather is none of the above.  It
does not claim homology with any part of a feather other than the collar 
and completely ignores the feather sheath.  It was created directly 
from the dinosaur protofeathers so there is no surprise that it looks like
them. 
 
I certainly don't believe the so called "feathers" in pterosaurs, and I am
amazed that nobody seems to to be upset with spreading feathers everywhere
especially when the majority of dinosaurs can be shown to be covered with scales. 

Why think that these things are feathers when they can't be distinguished
from fossilized muscles found in other organisms.
 
Notice that Mary's study comes from sediments where we might not expect
ancient preservation while they > are not duplicated where the opportunity
seems better. 

Prum's (1999) stage I feather is homologous to the calamus of modern
feathers, in the sense that it is formed by an undifferentiated follicle
collar.  Sorry for not specifying the calamus last time, I was grouping it
with the rachis in my mind.  Prum and Dyck (2003) describe sheath morphogy
and development in detail.
Though I can't claim to know Prum's thought processes, Prum and Dyck
describe how the model is largely dictated by the developmental mechanisms
of feathers (hierarchical modules), not merely resemblence to
Sinosauropteryx and Beipiaosaurus feathers.  Barbs have to evolve before
rachis because the rachis is just a fusion of barb ridges; barbules have to
evolve after barbs because they are formed by the differentiation of cell
layers in barb ridges; the calamus has to evolve first because the barb
ridges are differentiations of the collar that forms the calamus.  How do
you explain the structure of Longisquama's parafeathers being primitive for
feathers if they have a rachis and a barbless vane (as shown by those
successive magnifications at SVP 2001), when the rachis forms from fused
barb ridges?
 
Basically every pterosaur worker believes pterosaurs had filamentous
integument.  Why would you doubt it?  I know Feduccia cites Unwin and
Bakhurina (1994) as a contrary view, but they specifically noted such
integument was present in later papers and online communications.  They were
merely saying that the supposed Sordes hairs _figured up till that point_
were internal fibers.  I personally don't think they are feathers, but
that's only because there are so many scaled taxa between pterosaurs and
birds.  Filamentous integument evolved at least twice anyway (birds,
mammals), why not a third time?  And given the standard phylogeny, diapsid
filamentous integument only shows up twice- pterosaurs and coelurosaurs.
Unless you count Longisquama's or Psittacosaurus' structures. 

What specimens would these fossilized muscles be in?  In any case, they can
be distinguished from muscles and collagen (Lingham-Soliar, 2003) in the
following ways.
1. Dinosaur feathers are too long to be collagen fibers.  Even if one were
to hypothesize axial frills on Sinosauropteryx, how do you explain elongate
filaments on left and right sides of dorsoventrally preserved animals (NGMC
91; many Confuciusornis; Eoenantiornis)?  Elongate filaments on areas not
expected to have thick skin or need elongate muscle fibers (skull of
Microraptor; middle of Bewipiaosaurus' ulna)?
2. Dinosaur feathers are mostly perpendicular to the body wall (along ulna,
shoulders, femur of NGMC 91; head and chest of Microraptor; tail and ilium
of Sinosauropteryx), which is compatable with collagen, but not with muscle.
3. Dinosaur feathers are not inside the body wall like the fibers of
ichthyosaurs.  Ichthyosaurs have a very thick body wall, analogous to
cetaceans.  But proposing the body wall of theropods extended as far away
from the bones as seen in NGMC 91 or Eoenantiornis is preposterous.
4. Dinosaur feathers are hollow, unlike collagen.  Even if ichthyosaur
fibers look hollow due to mineralization stages (and you doubt Yixian
dinosaur feathers based on this), the filaments of Shuvuuia are three
dimensional, so cannot be explained in this manner.
5. Dinosaur feathers are suspiciously distributed similar to bird feathers.
There are none on the feet or distal snout.  Elongate feathers are
conveniently on the ulnar side of the lower arm (Beipiaosaurus, NGMC 91) and
distal tail (NGMC 91).  In particular, if they were muscles, we would expect
long tibial fibers on at least some specimens.
6. Dinosaur feathers are sometimes intermixed with vaned feathers, such as
on the head of Microraptor.  This is fairly indisputable proof that whatever
the more simple structures are, they are integument.
7. Dinosaur feathers are made of beta keratin, not collagen, actin or
myosin.  At least in Shuvuuia's case.
8. Why are Yixian lizards, choristoderes, salamanders and psittacosaurs not
covered with these structures?  They should have muscles and collagen, after
all.
How do you explain all of these factors?  I admit they could also be
homologous to turkey bristles (Sawyer et al., 2003), but the data seem to
argue against muscle or collagen.
 
Specimens from the Djadockhta can preserve keratinous structures, in the
form of claw tips (e.g. Citipati in Clark et al., 1999).  On the other hand,
how many specimens with preserved internal soft tissues have been found
there?
 
The theory that the rachis is fused barb ridges was rejected some decades
ago and the problems that it had then still aren't resolved. 
 
Why would you have a calmus if you have nothing else and where does the
feather sheath come in?  
 
Does Prum actually say that the protofeather is a calamus.  I don't
remember his saying so in any presentation.  I asked him once what part of a 
feather he was talking about, but the collar was all that I got out of it. 
You might try Paul Maderson if you want a critque on his model
of feather development.

Have you read Harris et al. (2002), which shows barb ridges fusing to form a
rachis in figure 4F-J?  Indeed, barb ridges can be induced to fuse to form
rachis if the follicle is split into two (Prum and Dyck, 2003).  Any
explanation?
Harris MK, Fallon JF, Prum RO. 2002. A Shh-Bmp2 developmental module and the
evolutionary origin and
diversification of feathers. J Exp Zool (Mol Dev Evol) 294:160-176.
 
Among the more plausible theories is for insulation (it would be comparable
to having fur).  Since the sheath is subsumed by the calamus once the latter
starts developing, and becomes an indistinguishable part of it, one might
assume a feather which is formed completely from the undifferentiated
'calamus setting' of the follicle collar would lack a sheath.  Sheaths would
have evolved only once the follicle collar became differentiated into barb
ridges. 

Prum says stage 1 feathers were formed by an undifferentiated follicle
collar, which is the same way that modern birds form their calamus.
Portions of feathers are only homologous in the sense that homologous
genetic patterns are being expressed at the time they are formed.
Maderson's model seems to contradict the developmental and genetic evidence
in asserting scales and feathers are homologous past the placode stage and
that barbs are formed from frayed scale edges.
 
Do you have any explanation for why muscle or collagen would exhibit the
eight traits I noted last time in coelurosaur and bird filaments?  They're
listed below in the copy of that message.
 
Comparison with dinosaurs should be through the common ancestor of the
maniraptorians and birds which according to the cladograms would have an 
avian wing.  If anything else is included as closer to maniraptorians 
than to birds, it would share that ancestor and hence be uninformative. 
 
If microraptor is a good model the common ancestor would not be bipedal. 

Longisquama has many feather features while protofeathers have essently none.
 
It also has an avian shoulder girdle. 

Yes, but tyrannosauroids and ornithomimosaurs would be outside the
maniraptoran+birds group, not necessarily have an avian wing (presence of
primaries unknown), and thus be informative.
 
Sure, but the stem leading to that common ancestor could be bipedal.  We
don't have Longisquama's hindlimbs, after all. 

I suppose I'll bite.  Like Longisquama's structures and bird feathers,
protofeathers are elongate, have a cylindrical base (hence, follicular
development), and a hollow shaft (Shuvuuia).  Pulp caps and sheaths are not
even seen in Yixian bird remiges or retrices, so their apparent absence in
protofeathers is meaningless.  The vanes of Longisquama's structures consist
of a two-layered sheet (since they were filled with sediment; Reisz and
Sues, 2001), so are only superficially similar to a feather vane (which
forms from barb ridges).  I fail to see any feather features parafeathers
have that are absent in protofeathers, let alone many.  Are there any I
missed?
 
Could you be more specific?  Coelurosaurs have slender scapulae and furculae
too, so those characters can't be used to place Longisquama closer to birds
than coelurosaurs.  I can't recall hearing of any other avian pectoral
characters in Longisquama.
 
Why you resist the one that has a lot of evidence for avian structure and
support the one that does not is more interesting to me than the debate. 
 
If you want to argue a bipedal ancestor and I would ask why you would even
care?  Than you must go quadrupedal to bipedal to quadrupedal to bipedal.  This seems to be
tedious and should only be pushed because of compelling evidence. 

If tyrannosaurs should turn out to be another secondarily terrestrial
lineage, I don't see how that advances the discussion.

It's exactly the same for me, though of course we have different views
regarding which specimens have a lot of evidence for avian structure and
which have less.  I don't expect to convince you my opinion is correct
(since it's been argued many times in the last couple decades by people with
far more experience) in this debate.  I'm mainly curious why you hold your
position and how you justify it.
But it's only fair I explain why I resist and support what I do too.  In
regards to Longisquama, here are the reasons I don't accept it as feathered.
Reisz and Sues (2000) have shown the supposed sheath is actually an
infilling of sediment.  This is proven by the fact that it covers the vane
corrugations on both part and counterpart, so must be internal.  I have not
heard a defense of why this is wrong from you or anyone else who supports a
feathered Longisquama.
The vane, being formed of two sheets, is unlike a feather vane in everything
except its expanded shape.  I've heard no developmental hypotheses for
changing this into a vane of barbs from you, Maderson or anyone else.  The
fact it can fold or is expanded distally has no relevence if it isn't
homologous.
Senter (2003) notes the supposed pulp caps are troughs between corrugations
on the middle vane.  He also describes the structures as being tripartite,
with anterior, middle and posterior vanes (the posterior vane narrows
distally until it disappears at midlength, while the middle and anterior
vanes expand to form the distal half.  I don't have the proper photos to
determine if these features are correct, but the fact multiple
interpretations exist certainly makes me weary of accepting Jones et al.'s
view (especially when the latter has proven incorrect in the ways described
above, and in regards to the supposedly separate barbs too- the successively
enlarged photos of the vane edge at SVP 2001 were convincing).
All this leaves in regards to featherlike characters is the narrow base,
which isn't too convincing since protofeathers exhibit it too, and the rest
of Longisquama's anatomy places it outside Sauria (Senter, 2003; 2004).
I've told you why I support protofeathers, with my eight ways they differ
from collagen and/or muscle.  Until you tell me why those are incorrect, we
can't progress on that topic.
 
When you said Microraptor was quadrupedal, I assumed you meant while
climbing.  While on the ground, you think Archaeopteryx and Confuciusornis
were bipedal, and since the former is so similar to Microraptor, I assumed
you thought it was bipedal on the ground too.  So if we are concerned about
terretrial locomotion only, birds are always bipedal, which is congruent
with ornithomimosaurs or tyrannosauroids being stem birds.  While climbing,
I don't see why small ornithomimosaurs or tyrannosauroids wouldn't use their
hands too.  Hell, with more arm/wrist mobility and perhaps no remiges, it
might have been easier for them than for maniraptorans. 

Here are my thoughts for how it could advance the discussion.
Foremost, I'm trying to understand how you justify separating maniraptorans
from theropods.  I know you'll use the homeotic shift for four-fingered
theropods, so I'm concentrating on the tridactyl ones.  You suggested
postacetabular length and supradentaries, but I showed those characters
don't work.  It's just such an odd concept to believe theropods and
maniraptorans are so distantly related, while being unable to assign some
taxa known from many complete skeletons to either group.  At least back when
you thought maniraptorans were dinosaurs, you had a list of characters that
could be used to separate dinosaurs from birds (hypopubic cup; expanded
tooth roots; four carpals; etc.).  So although you thought the groups were
superficially similar, these details could be used to differentiate them.
But that doesn't seem to be true for your new view.  This confuses me.
Secondly, your entire philosophy of 'not caring' about the relationships of
tyrannosauroids and ornithomimosaurs is confusing.  Finding that
maniraptorans are birds warrants a public talk, but what the next most
closely related group is seems unimportant to you.  There's a huge
morphological gap between Longisquama and birds, and you haven't been able
to give reasons tyrannosauroids and ornithomimosaurs don't fill that gap.
If they are stem birds, it would give us more evidence to use when deciding
what the next most closely related group is.  And I assume that question is
important to you.  This is analogous to my confusion regarding why you or
any other supporter of feathered Longisquama has never tried to figure out
where in Archosauria it goes.  The relationships between archosaurs have
been well resolved for over a decade, I would think such an important taxon
would be worth classifying more precisely.