Monday, December 15, 2014

Gansus zheni is Iteravis

So two Jehol ornithuromorphs were just named- Iteravis huchzermeyeri and Gansus zheni.  While writing up their Database entries, I noticed Iteravis is from the new Sihedang locality, where "most of the new ornithuromorphs appear to represent a single new taxon, which we describe here."  zheni is also from Sihedang.  Hmm...

How are the papers?

First a brief comment on the papers' quality.  Both taxa are named by different groups of researchers- Iteravis by Zhou, O'Connor and M. Wang, and zheni by Liu, Chiappe, Zhang, Bell, Meng, Ji and X. Wang.  Zhou et al.'s description of Iteravis is much longer and more detailed than Liu et al.'s on zheni, though Liu et al. provide more closeups of the specimens.  One irritating thing about Liu et al.'s measurement table is that it excludes manual elements except for the carpometacarpus.  But it does waste space with previously published measurements of Gansus yumenensis.  And I just noticed now that Gansus' name is put above the wrong column, the one for the right side of zheni specimen BMNHC-Ph 1342.  Good thing we pay journals to typeset for us!  Also annoying is both papers mention specimens that are undescribed.  Zhou et al. mention "approximately 20 ornithuromorphs" from Sihedang, of which most are Iteravis, but which don't even get specimen numbers.  Liu et al. list BMNHC-Ph 1394 as a paratype of zheni, but its not illustrated, measured or described.  While there are a few things I think Zhou et al. got wrong, Liu et al. make more mistakes in both anatomy and grammar.  The main skeletal figures in Liu et al. (figures 1 and 2) have the specimen numbers switched, so figure 1 says its of BMNHC-Ph 1342 but is actually of BMNHC-Ph 1318, and the reverse is true of figure 2.  On the other hand, Zhou et al. claim they used the data matrix of O'Connor et al.'s (2011) redescription of Rapaxavis, but that paper has no phylogenetic analysis.  As far as I can tell, they actually used the matrix of O'Connor and Zelenkov's (2013) redescription of Ambiortus, which isn't even in their References list.  Both of these are pretty big blunders you would hope peer reviewers would catch.  In the analysis itself, Zhou et al. initially found one most parsimonious tree in TNT, then ran it a second time (with that first tree as the seed?) and found a whopping 9760 trees that were one step shorter.  They show both that first 847 step tree and the strict consensus of the 846 step trees, but why bother showing the less parsimonious tree that's misleadingly resolved?  And if you have a huge polytomy involving Iteravis and eight other taxa in your strict consensus, why not make any attempt to resolve it?!  It's VERY easy in TNT to remove any combination of taxa from your consensus tree and thus see which ones cause the polytomy.  My guess this time is that the fragmentary Limenavis is messing things up.  I'd check, but the authors didn't include their codings for Iteravis in the paper and there's no supplementary information.  Guess I'll be emailing Jingmai...

Zhou et al. think Sihedang is in the Yixian Formation and Liu et al. think it's in the Jiufotang Formation, and the latter seem correct as the pterosaur Ikrandraco is known from Sihedang and another Jiufotang locality.  The birds are near identical in size (femora 35 vs. 34.5-36.4 mm) and have similar morphology.  Let's compare diagnoses-

Iteravis' diagnosis

Iteravis is diagnosed by (after Zhou et al., 2014)-
A1. premaxillary corpus elongate and toothless. The premaxillary body is the same length in zheni, and lacks teeth there as well (fig. 3A, contra Liu et al. stating it is unclear whether teeth were present; see red line in my cranial figure).  Toothlessness in the premaxilla is actually a synapomorphy of birds closer to Aves than songlingornithids. 
A2. maxilla with numerous teeth.  Again stated to be uncertain in zheni by Liu et al., but five alveoli are visible in their figure 3A (see red angle in my cranial figure).  As for the tooth number, Zhou et al. merely report "several" in Iteravis, and only a couple are visible in their figure, so this matches as well.  Ornithuromorphs plesiomorphically have at least several maxillary teeth.
A3. rostrum 50% of skull length. This is also true in zheni, and in many other basal ornithuromorphs too- Archaeorhynchus, Hongshanornis, Jianchangornis, Schizooura, Longicrusavis, Yixianornis, etc..
A4. ethmoid bone lining rostral half of the orbit. Though unidentified in zheni, this is visible in figure 3B and even has the same posterodorsally angled ridge with expanded central portion (yellow lines in cranial figure).  Of course, this mesethmoid is widespread in maniraptoriforms, though rarely preserved.
A5. tubercle on caudal margin of minor digit phalanx. This is visible in zheni too (green line in my postcranial figure), though also in Yumenornis, Gansus (Li et al., 2011 fig. 2A) and Ichthyornis.  Clarke (2004) states this flexor process is also present in tinamous "and an array of other avian taxa", and as phalanx III-1 isn't preserved in other non-avian birds more derived than songlingornithids, it may be a character of this clade.
A6. pubes with dorsally expanded distal boot. The pubic boots are almost identical in Iteravis and zheni.  Boots that are dorsally/posteriorly expanded are plesiomorphic for theropods, and are found in other ornithuromorphs that retain boots such as Hongshanornis, Yanornis and Schizooura.
A7. narrow ischium with concave ventral margin and weak dorsal process at midpoint.  zheni has the same kind of ischium.  A number of ornithuromorphs have ventrally concave ischia (e.g. Chaoyangia, Schizooura, Changmaornis, Yixianornis, hesperornithines), and of these, Gansus, Piscivoravis and Yanornis have a weak mid dorsal process as well.

Skulls of Iteravis zheni. Top inset- Gansus zheni holotype BMNHC-Ph 1342. Top- Gansus zheni paratype BMNHC-Ph 1318. Bottom- Iteravis huchzermeyeri holotype IVPPV18958. Red line points to toothless premaxillary edge, supposedly uncertain in Gansus zheni.  Red angle shows possible maxillary alveoli.  Yellow line points to mesethmoid ridge similar in both specimens.  Blue line points to predentary, supposedly unpreserved in Gansus zheni. Modified after Liu et al. (2014) and Zhou et al. (2014).

zheni's diagnosis

zheni is diagnosed by Liu et al. purely by characters that supposedly distinguish it from Gansus yumenensis-
B1. broader U-shaped furcula with an interclavicular angle of approximately 60 degrees (as opposed to ~40 degrees for Gansus yumenensis).  This is untrue, as interclavicular angles in zheni are ~45 and ~53 degrees.  Iteravis' is 54 degrees and Gansus' is ~42 (listed as ~38 by You et al., 2006).  The numbers for zheni and Gansus should be taken with caution though, as their furculae are distorted.  Similar angles to Iteravis and zheni are found in Archaeorhynchus, Jianchangornis, Parahongshanornis, Hongshanornis, Schizooura, Songlingornis and Yixianornis.  Almost all basal ornithuromorphs have U-shaped furculae.
B2. cnemial crests of tibiotarsus extending distally less than in Gansus yumenensis.  This is untrue, as Gansus has a very short cnemial crest extending 4% of tibiotarsal length (excluding the apomorphic proximal extension of the crest), while zheni's cnemial crest is 15% of tibiotarsal length (not 25% as the authors state).  Iteravis' is 8% (comparable to the authors' estimate of 10%).
B3. manual digit II (major digit) proportionally shorter than in Gansus yumenensis.  Phalanges II-1+2 are 86-96% of metacarpal II in zheni compared to 102% in Iteravis.  This is compared to 81-83% in Gansus, so the character is incorrect.  Similar values to Iteravis are found in Tianyuornis (85%), Yumenornis (87%), Piscivoravis (88%), Schizooura (93%), Archaeorhynchus (91-96%) and Yanornis martini (95-107%).
B4. pedal digit IV slightly longer than digit III (pedal digit IV is markedly longer than digit III in Gansus yumenensis). The ratio (ignoring unguals) in zheni is 99-106% compared to 110% in Iteravis.  The ratio in Gansus is 110-122%, so is usually larger, but not always distinct.  Schizooura is the only other basal ornithuromorph which overlaps Iteravis/zheni, with a ratio of 100%.
B5. ratio of pedal digit III/tarsometatarsus about 1.1 (it is approximately 1.0 in Gansus yumenensis). Ignoring unguals (which are rarely measured), the ratio is 89-97% in Iteravis and 97% in zheniGansus actually overlaps this with ratios of 74-101%.  Several other basal ornithuromorphs also overlap Iteravis, such as Jianchangornis (97%), Piscivoravis (94%), Yanornis martini (94-96%) and Y? guozghangi (97%). 
B6. claws of pedal digits III and IV lacking a prominent pendant flexor tubercle (present in Gansus yumenensis).  This is also true in Iteravis and all other basal ornithuromorphs, as the flexor tubercle morphology of Gansus is an autapomorphy.

Skeletons of Iteravis zheni. Left- Gansus zheni paratype BMNHC-Ph 1318. Center- Gansus zheni holotype BMNHC-Ph 1342. Right- Iteravis huchzermeyeri holotype IVPP V18958. Yellow line points to supposed median ventral synsacral groove which is more likely taphonomic.  Green line points to flexor process on manual phalanx III-1. Purple line points to omal process on lateral coracoid, supposedly absent in Iteravis (note the impression in the matrix).  Modified after Liu et al. (2014) and Zhou et al. (2014).
Other supposed differences

You can see that given their diagnoses, Iteravis and zheni only differ in slightly different ratios that are usually closer to each other than between the two measured zheni specimens.  There are also several characters which differ in their descriptions.   

zheni is said to have a "small, rostrally tapered, and tear-shaped" external naris (mistakenly cited as the internal naris), but given the odd premaxillary shape in BMNHC-Ph 1318, the premaxilla and maxilla are probably crushed in largely ventral view (note several possible alveoli in the maxilla and the deep bone under them which would be the palatal shelf; see red angle in my cranial figure), artificially shortening and tapering the anterior narial edge.  Liu et al. state zheni's naris posteriorly overlaps the antorbital fenestra, which would barely be true in their interpretation, while the labeled nasal fragment in Iteravis suggests this isn't so in that taxon.  However, the antorbital fenestral area in both specimens is a jumble of bone fragments and multicolored sediment reflecting the fragile nature of that region in birds and the separation of slabs which exposed it.  Thus any edge of the fenestra is impossible to identify exactly.  Liu et al. claim "Unlike other Jehol ornithuromorphs [including Iteravis] ... no pre-mandibular ossification is visible in any of the two studied specimens."  This would be easily explainable by taphonomy as both skulls are rather poorly preserved and the element is small and loosely connected to the dentaries.  Regardless, there are possible predentaries in each specimen- contacting the premaxillae just in front of the dentary in BMNHC-Ph 1342 and attached to the left dentary tip projecting dorsally in BMNHC-Ph 1318 (blue lines in my cranial figure).

Liu et al. state zheni has "a broad ventral groove running along the entire exposed surface" of the synsacrum, while Zhou et al. state Iteravis has "a flat ventral surface".  The latter seems true, but the 'groove' in zheni seems to be the taphonomic collapse of the hollow interior as seen in its tibiotarsi, humeri and ulnae (yellow line in my postcranial figure).  Zhou et al. states Iteravis lacks "the cranial hook present in Gansus", while it is clearly present in zheni's coracoids.  Yet both coracoids are broken in this area in Iteravis, and the left shows a depression in the matrix which seems to indicate the hook's original presence (purple line in my postcranial figure).  Liu et al. state zheni has a "prominent and triangular-shaped laterocranial process", which is absent in Iteravis.  Yet this process is also absent in the illustrated zheni specimens BMNHC-Ph 1318 and 1342.  Liu et al. cites BMNHC-Ph 1394 as having the process, but until this specimen is illustrated, it can be considered polymorphic at best to misinterpreted at worst.  zheni is said to lack ossified uncinate processes, whereas Iteravis is reported to preserve "one probable uncinate process".  All three specimens have ribcages which are only partly articulated and exposed though, so its easily possible uncinate processes are hidden if present in zheni, or that the one was misidentified in Iteravis.  Liu et al. state zheni has a deltopectoral crest on the humerus "which extends more than one-third the total length of the bone", while Iteravis' is described as extending "the proximal one-third of the humerus", but the crest in the latter is almost entirely covered by other elements so cannot be measured.  Iteravis' carpometacarpus is described as incompletely fused versus completely fused in zheni, but the specimen is slightly smaller than zheni specimens (humeri 97% of BMNHC-Ph1318, 95-97% of BMNHC-Ph1342) so could be expected to have less fusion.  Liu et al. say zheni lacks an extensor process on metacarpal I, while Iteravis is said to have a small extensor process.  Both taxa have the same morphology though, which is comparable to the extensor flange of basal paravians and not the extensor process of some ornithuromorphs.

As noted above, the authors give very different lengths for Iteravis' and zheni's cnemial crests (10 vs. 25% of tibiotarsal length), though the real apparent values are 8% vs. 15%.  The discrepancy largely seems due to zheni's tibiotarsi being preserved in anterior view, where the collapse of the element causes a median groove that exaggerates structures on either side such as the laterally placed cnemial crest.  Iteravis' right tibiotarsus is in medial view, but the left element is partially covered by the sternum and has a taphonomic concavity that extends the apparent length of the cnemial crest.  Iteravis' fibula is described as "just over half the length of the tibiotarsus", while zheni's is said to only extend "to nearly the midshaft of the tibia."  In reality, all specimens have distal ends hidden by the tibiotarsus so cannot be exactly measured.  Liu et al. state "the proximal phalanges of all pedal digits are longer than any of their respective distal phalanges" in zheni, while Zhou et al. say Iteravis has a slightly longer II-2 than II-1.  Their own measurement table shows zheni is polymorphic for this though.  Iteravis is reported to have a pedal digit IV shorter than III in contrast to zheni, but as noted above, the ratio excluding unguals is 110% in Iteravis vs. 99-106% in zheni.  So Iteravis actually has the longer digit IV, but there's more variation in zheni than difference between it and Iteravis.

Given the lack of difference between Iteravis and zheni, they are near certainly synonyms.  Iteravis was published online October 29th vs. zheni on November 14th.  We might think huchzermeyeri wins over zheni by 16 days, except for the fact Zhou et al. didn't include a ZooBank registration.  So the physical publication time is what counts, which is December 1. Thus zheni wins by 30 days.  A good example of why people need to register their names. Still we have the question...

Is this Gansus?

Liu et al. diagnosed Gansus by-
C1. hooked omal projection of the sternolateral process of the coracoid. This is actually polymorphic in Gansus (absent in CAGS-IG-04-CM-012; O'Connor and Zelenkov, 2013), and also present in Ichthyornis.  Both are miscoded in the matrix.
C2. sternum with a caudomedially curved outermost (lateral) trabecula. This is more accurately understood as having a distal expansion of the posterolateral process that is expanded medially but not much laterally, which makes the end seem to curve medially.  It is also present in Jiuquanornis, Hongshanornis, Jianchangornis, Yumenornis and Ambiortus. Notably, this isn't actually the character coded for in the analysis, which is the angle the process projects at.
C3. intermembral index (length of humerus + ulna)/(length of femur + tibiotarsus) between 0.9 and 1.1. zheni actually falls outside of this range sometimes, with a ratio of 1.06-1.12.  Parahongshanornis (.9) and Yixianornis (1.07) also fall within this ratio.
C4. intermetacarpal space terminating distal to the distal end of metacarpal I.  This is unreliable, as it varies with metacarpal I length as well as how the laminar metacarpal III is crushed in relation to metacarpal II.  
C5. metatarsal II shorter than metatarsal IV (metatarsal II extends distally only as far as the base of the trochlea of metatarsal IV). Metatarsal II is near universally shorter than IV in basal ornithuromorphs, with their exact trochlear relation difficult to judge in specimens preserved with tarsometatarsi in dorsal view. However, metatarsal II is as short compared to IV as in Gansus or moreso in Ichthyornis and hesperornithines at least.  This is coded as such in the matrix, but the basal hesperornithines Enaliornis and Baptornis are miscoded as having longer mtIIs, as is Vegavis.  With these corrected, all included birds more derived than songlingornithids have the state, except Apsaravis and Gallus.
C6. proximal pedal phalanges longer than distal pedal phalanges. This is true in almost every basal ornithuromorph, with zheni and Gansus ironically being the only taxa with some discordant specimens (both in digit II).  Note while this is a character in the analysis, it is properly coded as present in all ornithuromorphs in the matrix so does not resolve as a Gansus character.  Why it was then included in their emended diagnosis for the genus, I have no idea.
C7. pedal digit IV longer than digit III. zheni is polymorphic for this, as noted above.  Hesperornithoids also exhibit the state, more developed than zheni or Gansus.  The character was not included in their analysis.
They also listed one other character which united zheni and Gansus in their analysis-
C8. supracoracoidal nerve foramen of coracoid displaced so that it[s] nerve no longer passes through the coracoid. Zhou et al. state this is untrue in Iteravis. The foramen is difficult to see in specimens crushed in two dimensions, so it is very possible Liu et al. miscoded zheni.  Or it may be polymorphic.  Archaeorhynchus, Jianchangornis, Patagopteryx, Hongshanornis and Longicrusavis also lack this foramen, but only the former two are coded as such in the matrix.

The total evidence here suggests C4 is invalid, C1, C3, C6, C7 and ?C8 are polymorphic in at least one those species, C5 is symplesiomorphic, and C2 is true in several other taxa.  Changing the ten miscodings noted above (with zheni conservatively coded as polymorphic for the coracoid foramen), zheni ends up basal to Gansus and birds closer to the crown.  Checking which characters supported this, I found most were based on miscodings as well*.  Correcting these left zheni basal to Gansus, supported only by its gastralia.  Yet Gansus specimens may have taphonomically lost their gastralia (e.g. no crania are connected to any), so this isn't the greatest evidence.  Enforcing zheni to be Gansus results in trees one step longer, so is basically as parsimonious.  Thus neither position is well supported, and the new combination Iteravis zheni is used until good evidence for referring it to Gansus is presented.

* Ichthyornis and Apsaravis miscoded as having fused dentaries; Gansus miscoded as having an extensor process on metacarpal I; Gansus miscoded as having a first manual digit shorter than 50% of metacarpal II length; Gansus miscoded as having a manual ungual II subequal to or longer than I; Schizooura, Gansus, Ichthyornis, Anas and Gallus miscoded as having proximodorsal ischial processes, when they have mid-dorsal processes like other ornithuromorphs (except Patagopteryx which has both and was miscoded as lacking the proximodorsal one); zheni miscoded as having 8 and/or 10 sacrals instead of ~8-10.

Finally, the 22 miscodings noted above change the ornithuromorph part of the tree.
Liu et al. had- (Arc(Pat(Vor(Schiz((Hong,Long)(Jian((Song(Yan,Yix))(Gans(Hesp(Ich(Veg,Aves)))))))))))
... with Apsaravis either just more derived than Gansus or sister to Vegavis+Aves.
The new tree is- ((Vor,Pat)(Schiz(Arc,Jian,Song,Yan,Yix(Long,Hong)(Gans(Ich(Aps(Hesp(Veg,Aves))))))))
Not that I trust the latter tree either, as 22 corrected states centered around one node changed that much, further corrections will probably have greater effects.

References- Clarke, 2004. Morphology, phylogenetic taxonomy, and systematics of Ichthyornis and Apatornis (Avialae: Ornithurae). Bulletin of the American Museum of Natural History. 286, 1-179.

You, Lamanna, Harris, Chiappe, O'Connor, Ji, Lu, Yuan, Li, Zhang, Lacovara, Dodson and Ji, 2006. A nearly modern amphibious bird from the Early Cretaceous of Northwestern China. Science. 312, 1640-1643.

Li, Zhang, Zhou, Li, Liu and Wang, 2011. New material of Gansus and a discussion on its habit. Vertebrata PalAsiatica. 49(4), 435-445.

O'Connor, Chiappe, Gao and Zhao, 2011. Anatomy of the Early Cretaceous enantiornithine bird Rapaxavis pani. Acta Palaeontologica Polonica. 56(3), 463-475.

O'Connor and Zelenkov, 2013. The phylogenetic position of Ambiortus: Comparison with other Mesozoic birds from Asia. Paleontological Journal. 47(11), 1270-1281.

Liu, Chiappe, Zhang, Bell, Meng, Ji and Wang, 2014. An advanced, new long-legged bird from the Early Cretaceous of the Jehol Group (northeastern China): Insights into the temporal divergence of modern birds. Zootaxa. 3884(3), 253-266.

Zhou, O'Connor and Wang, 2014. A new species from an ornithuromorph (Aves: Ornithothoraces) dominated locality of the Jehol Biota. Chinese Science Bulletin. 59(36), 5366-5378.


  1. Good catch. This is one of your blog posts that probably deserves to be an actual paper.

  2. I have to score the new Gansus species in my dataset, but agree that it may not be a Gansus and eventually closely related (if not the same taxon) as Iteravis. My analysis provisionally places Iteravis more basal than Gansus:

  3. Good catch. This is one of your blog posts that probably deserves to be an actual paper.

    Seconded and thirded.

  4. Hi Mickey,

    In their description of the basal ornithuran specimen AGB5841, the oldest fossil bird with a supraorbital salt gland fossa, Wang et al. (2018) agree with your conclusion that Gansus zheni is more closely related to Iteravis than to the Gansus type species based on shared characters between AGB5841, IVPP V18958, and BMNHC-Ph 1342 and the results of a cladistic analysis.

    Xia Wang; Jiandong Huang; Yuanchao Hu; Xiaoyu Liu; Jennifer Peteya; Julia A. Clarke (2018). "The earliest evidence for a supraorbital salt gland in dinosaurs in new Early Cretaceous ornithurines". Scientific Reports. 8: Article number 3969. doi:10.1038/s41598-018-22412-8.