[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]
Our current understanding of Mesozoic bird phylogeny
David's recent Gansus question reminded me of this study I did a couple
months ago. And yes, his question gets answered towards the end.
While updating my website, I noticed the lack of rigorously tested
phylogenies for basal birds, especially enantiornithines. While my ongoing
analysis will eventually answer this question to my satisfaction, I wanted
an idea of their relationships now, as opposed to months in the future. So
over the last few weeks, I've hastily combined Clarke's (2002, 2004, etc.)
matrix with Chiappe's (2001, 2002, etc.) and Chiappe et al.'s (2006), coded
almost all relevent taxa, and added several more characters from other
sources. This includes all characters that have been used in the past for
numerical analyses of enantiornithine interrelationships, and most that have
been used for bird relationships at levels between Avialae and Neornithes.
The downsides are that I didn't check my or the authors' codings as
rigorously as I do for my big analysis, so miscodings are no doubt present.
Also, a lot of characters are divided into coarser states than need be,
since that's how nearly every one is defined in the literature at this time.
This is the current understanding of bird phylogeny, not the future
What characters did I include?
1-202 of Norell and Clarke (2001), as used subsequently in Clarke and Norell
(2002), Clarke's (2002) thesis, her (2004) Ichthyornis paper, the new
Yixianornis paper (Clarke et al., 2006) and the descriptions of Jeholornis
(Zhou and Zhang, 2002), Sapeornis (Zhou and Zhang, 2002), Jinfengopteryx (Ji
et al., 2005), Hongshanornis (Zhou and Zhang, 2005), and Archaeorhynchus
(Zhou and Zhang, 2006). Clarke et al. (2006) had four extra characters that
were also included.
The 19 characters from Chiappe et al. (2006) enantiornithine analysis in the
Elsornis paper which weren't used in the above.
Three characters from Chiappe and Walker's (2002) euenantiornithine analysis
which weren't used in the above.
The 37 characters from Chiappe (2001, 2002) which weren't present in the
above and weren't only used for inter/relationships of alvarezsaurids.
Two characters of my own (median sternal process expanded distally; mtIV
longer than mtIII) that I thought might affect enantiornithine
I also altered some of these characters. Cervical centrum morphology is
only evaluated for anterior cervicals, since in several taxa posterior
cervicals are still amphicoelous when anterior ones are semiheterocoelous.
Dorsal vertebral count was divided so that 12 and 13 were different states.
Dorsal parapophysis placement was only evaluated in posterior dorsals, since
anterior dorsals can maintain the plesiomorphic anterior placement in some
taxa (e.g. Iberomesornis). Caudal vertebral count was revised to be a total
count and not just of free vertebrae. I quantified the ulnohumeral ratio to
be more objective. I divided manual phalanx II-2/II-1 ratios into more
states as in my analysis. I ordered posterior trochanter size. I
quantified and added a state to tibiotarsal medial/lateral condyle width
ratios. I separated distal tarsal fusion from metatarsal fusion. I made
each possible number of phalanges in manual digit III its own state.
Characters were ordered where morphological gradients existed.
What taxa did I exclude?
I excluded some non-pygostylian taxa like Wellnhoferia, Jixiangornis,
Yandangornis, Dalianraptor, Omnivoropteryx, as well as controversially basal
bird taxa like Hulsanpes, Jinfengopteryx, Unenlagia, Buitreraptor, Rahonavis
and scansoriopterygids. Any useful analysis of these taxa would have to
include other paravians and is outside the scope of this analysis. The
analyses I used for characters were largely designed to test relationships
within Pygostylia, so Sapeornis and Shenzhouraptor (=Jeholornis) are decent
outgroups, along with Archaeopteryx and a Dromaeosauridae that could no
doubt stand to be coded better. Also excluded were taxa which probably fall
inside Pygostylia but cannot be coded for more than a couple characters
(endocast- Cerebavis; sacrum- Kuszholia, Lenesornis, Platanavis, Zhyraornis,
Z? logunovi, Guildavis, Gargantuavis; distal radius- "Paleopteryx"; feather-
Cretaaviculus, Illerdopteryx, Praeornis; undescribed or illustrated-
"Holbotia"). I don't have the descriptions or good illustrations of
"Cathayornis" aberransis, Eurolimnornis or Palaeocursornis, so these were
excluded as well. Neither Horezmavis nor hesperornithines besides Baptornis
and Hesperornis were included, though I could try this in the future.
Everything else (66 taxa) has been included, plus 5 unnamed specimens.
Finally, Jinzhouornis wasn't included, as I had previously concluded the two
species are junior synonyms of Confuciusornis, and probably C. sanctus
I ran the analysis with all taxa except Dapingfangornis and those described
by Hou (1997) - Jibeinia, "Cathayornis" caudatus, Longchengornis,
Cuspirostrisornis, Largirostrornis. This is because Hou's descriptions of
other taxa in that work (Confuciusornis, etc.) are inaccurate and the
illustrations are of very low quality. For instance, Confuciusornis is
described as having a septomaxilla, prefrontal, patella and a free distal
tarsal, contra Chiappe et al. (1999). Because every taxon affects every
other taxon's placement, I didn't want inaccurate codings for Hou's taxa to
change character distribution and topology of the whole tree.
Dapingfangornis has a similarly bad illustration, and the authors'
interpretation of the skull at least seems to be flawed (see
http://dml.cmnh.org/2006Apr/msg00170.html). Each of these taxa was
subsequently added individually to test for its placement in the cladogram,
but these must be viewed with caution due to the numerous probable
Of course the large amount of missing data resulted in huge polytomies, but
it was important to include even fragmentary taxa (as long as they had
unique character combinations) in order to get a tree with the best
character distribution and topology for all taxa. Fragmentary taxa were
then excluded from the tree (but not the analysis) until a fully resolved
tree was present. Each taxon was then reintroduced to the tree separately
to determine the possible places it could go in the phylogeny.
The resulting cladogram is shown below. It uses the same format as my
website, with an asterisk meaning the taxon is incertae sedis within the
node it's placed at. Most clade names have been phylogenetically defined,
with the exception of the enantiornithine families and subfamilies (except
`--Ornithurae sensu Gauthier
| |*-Explorornis nessovi
| `--+*-Explorornis? walkeri
| | |--Noguerornis
| | `--+--Protopteryx
| | `?-Jibeinia
| |*?"Cathayornis" caudatus
| | |--Longipteryx
| | `--Longirostravis
| |--IVPP V14238
| |?-Jibeinia adult
| | `?-Dapingfangornis
| | |--Hebeiornis
| | |--LP-4450-IEI
| | `--+--GMV 2158
| | `--GMV-2159
| | |--Boluochia
| | `--Gobipteryx
| | |--Avisaurus
| | `--Soroavisaurus
| | |--Eoalulavis
| | `--Liaoningornis
`--Ornithurae sensu Gauthier and deQuieroz
`--Ornithurae sensu Chiappe
I initially wanted to include diagnoses for the clades found here, but they
would make the post far too long. So diagnoses will be on my website once I
update it with this phylogeny. The tree above (without Hou's taxa and
Dapingfangornis) is 857 steps long. I examined a number of proposed
alternative phylogenies to determine how much longer they were, giving a
rough idea of their probability.
895 steps- Patagopteryx as a palaeognath
This was the original position proposed for Patagopteryx by Alvarenga and
Bonaparte (1992) and also by Kurochkin (1995), though Chiappe rejected it as
early as 1989.
892 steps- Confuciusornithids as euornithines and archaeopterygids as
This was suggested by Kurochkin (2006), whose paper I critiqued here-
http://dml.cmnh.org/2006Oct/msg00467.html . Of course, since he also has
all non-bird theropods more closely related to sauriurines than to
euornithines, his phylogeny is even less parsimonious than can be shown here
(as non-bird theropods besides dromaeosaurids are not included).
891 steps- Archaeopteryx and confuciusornithids as sauriurines
This goes back to Hou et al.'s (1995) description of Confuciusornis, and is
still supported by MANIACs like Martin and Feduccia.
880 steps- Sinornis outside of Enantiornis + Passer
This kind of paraphyletic Enantiornithes was found by Chatterjee (1999) and
also supported as a possibility by myself over the years, but seems not to
be true. In Chatterjee's case, it was due to a large number of incorrect
875 steps- Otogornis as a euornithine ambiortiform
This is Kurochkin's (1999) idea, based on his reinterpretation of the
material. Even if I were to grant Kurochkin's alternative morphology
(coracoid tubercle = procoracoid process), the tree would still be 873 steps
or so, showing Kurochkin's identifications for these are probably wrong.
874 steps- Liaoningornis as a euornithine
This was proposed by Hou (1997) and still followed as recently as Zhou and
Zhang (2006). As I noted in http://dml.cmnh.org/2005Dec/msg00237.html , the
cladistic disagreements are largely due to different interpretations of
morphology. I used Clarke's (2002) codings here.
874 steps- Hebeiornis as a protopterygid
Remember that Hebeiornis is the correct name for Vescornis, as I detailed
here- http://dml.cmnh.org/2006Dec/msg00079.html. Zhou and Zhang (2006)
placed it together with Protopteryx and Jibeinia in the Protopterygidae,
871 steps- Protopteryx and Longipteryx successively closer to
This kind of paraphyletic Enantiornithes was proposed by me back in 2001 on
the DML, based on six characters, all of which were examined here. Notably,
far more enantiornithine characters have been described in it now than were
proposed by Zhang et al. (2001). I now officially reject this hypothesis.
871 steps- Patagopteryx as a hesperornithine
Chatterjee (1999) found this result, which was based on several characters,
some miscodings and others due to reduced forelimbs.
870 steps- Patagopteryx outside of Ornithothoraces
I've noticed several characters that are remarkably basal in Patagopteryx,
such as the unreduced proximal coracoid and brevis shelf, but apparently
these are near certainly reversals.
Not likely, but possible
867 steps- Iberomesornis outside of Enantiornithes + Euornithes
This idea was popular prior to Sereno (2000) and Chiappe (2001), due largely
to several misinterpretations of Iberomesornis. Indeed, Sanz et al. (2002)
continue to make some of them- five sacrals, unfused astragali and distal
tarsals, unfused metatarsals.
867 steps- Euornithiformes sensu Kurochkin 1996
Kurochkin proposed a basal clade of Enantiornithes containing Iberomesornis,
Noguerrnis, Concornis, Sinornis and Boluochia, but excluding the Lecho taxa,
Alexornis, Gobipteryx, Nanantius, Kizylkumavis and Sazavis. This was based
on seven characters, most of which are vague and flawed (as in his other
enantiornithine clades, see here-
http://dml.cmnh.org/2001Feb/msg00618.html). Euornithiformes fails largely
due to the avisaurid characters in Sinornis and Concornis.
867 steps- Gansus as a hesperornithine
This has never been suggested in the literature, but I was curious how
unparsimonious grouping these aquatic taxa together would be.
867 steps- Ambiortus as a palaeognath
This goes back to Kurochkin (1985), and was still entertained by him in 1999
865 steps- Chaoyangia as a confuciusornithid
Though near universally identified as a euornithine (=ornithurine of older
workers), Clarke (2002) found that Chaoyangia grouped with Confuciusornis
instead and might even be synonymous with that taxon. The analysis of more
characters makes this unlikely.
865 steps- Protopteryx outside of Ornithothoraces
This idea was developed by me on the DML in 2000. It was based on eleven
characters, nine of which were included in this analysis. One of the others
(maxillary fenestra) is now known in some enantiornithines anyway. Of the
included characters, some are now known in basal euornithines (two phalanges
on manual digit III) or derived enantiornithines (gastralia; extensor
865 steps- Alexornithiformes sensu Kurochkin 1996
This is the sister taxon to Kurochkin's Euornithiformes, with the opposite
taxon distribution than described for the latter clade. It is slightly more
likely, as alexornithiform taxa are more poorly known and generally more
865 steps- Aberratiodontus as an enantiornithine
This was suggested in the original description by Gong et al. (2004), but
wasn't supported by many characters. The few mentioned were plesiomorphies
(elongate pygostyle; posterolateral sternal processes; narrow posteromedian
process; unfused carpometacarpus; distally unfused tarsometatarsus) or
misinterpretations (well developed postorbital; short sternum). The
addition of several euornithine characters makes the taxon appear as a basal
member of that clade, but a more detailed description could certainly change
865 steps- Ambiortus as an ichthyornithine
This is an idea proposed by Martin (1987), though not supposed in any
explicit cladistic analysis published to date.
863 steps- Otogornis sister to Longirostravis
This was proposed by Zhou and Zhang (2006) without justification. It is
possible primarily due to the fragmentary nature of Otogornis.
862 steps- confuciusornithids as enantiornithines
Surprisingly, this reduced version of Sauriurae (which shouldn't be termed
Sauriurae, as it lacks Archaeopteryx) is highly parsimonious. The addition
of Chiappe's characters makes it slightly less so than the traditional
(western) ornithothoracine view.
862 steps- Chaoyangia as an enantiornithine
This was found by Clarke (2002) to be equally parsimonious to topologies
where it was a confuciusornithid. When more characters are added, this
possibility becomes more likely and almost as likely as those in which it is
861 steps- Noguerornis as an iberomesornithid
This was proposed by Chiappe (2001) based on one character- a non-procoelous
sacrum. Similarly, the present position sister to Protopteryx is only based
on the elongate ulna in both taxa. Noguerornis' relationship is hard to pin
down, though it does seem to be a relatively basal enantiornithine.
861 steps- Alexornithidae sensu Kurochkin 1996
Kurochkin proposed a family uniting Neuquenornis, Alexornis, Gobipteryx,
Nanantius, Kizylkumavis and Sazavis to the exclusion of the Lecho taxa and
Early Cretaceous enantiornithines he was aware of. As these taxa are
especially poorly known (except for Gobipteryx and Neuquenornis), they can
easily switch positions to form a clade. However, Kurochkin's proposed
synapomorphies for this clade are almost entirely invalid, as shown in my
post linked to above.
861 steps- Nanantius as a gobipterygid
Kurochkin (1996) felt that the most complete specimen of Gobipteryx was in
fact a new species of Nanantius- N. valifanovi. This was based on
tibiotarsal characters that are mostly not included in this analysis, nor
possible to compare with most enantiornithine taxa. The present position as
an iberomesornithid is only due to the completely fused tibiotarsus, which
is possibly ontogenetic. A more derived position for Nanantius, such as one
in Gobipterygidae, is therefore quite possible.
861 steps- Enantiornithidae sensu Kurochkin 1996
Kurochkin's Enantiornithidae contains Avisaurus, Soroavisaurus and
Enantiornis. While the first two taxa are closely related, Enantiornis
cannot be compared to either of them, being known only from forelimb and
pectoral elements. Although Enantiornis is here excluded from Avisauridae
due to the non-compressed distal humerus, it doesn't take many extra steps
to make it an avisaurine.
861 steps- Neuquenornis as an avisaurine
Sanz et al. (1995), based on Chiappe (1993), conducted a phylogenetic
analysis of enantiornitines based on pedal characters, with the topology
(Lectavis,Yungavolucris (Concornis (Neuquenornis (Soroavisaurus,
Avisaurus)))). This agrees with the present analysis except that
Neuquenornis is an avisaurine instead of a concornithine.
861 steps- Chaoyangia sister to Songlingornis
Originally (Hou et al., 1996), Songlingornis was described as a specimen of
Chaoyangia. The two holotype are not comparable though, besides some
fragmentary portions. This position is less parsimonious than its current
one sister to Archaeorhynchus mainly because Chaoyangia has a low number of
sacrals and lacks a completely fused pelvis.
861 steps- Odontornithes
Another surprisingly likely possibility is that Hesperornithes and
Ichthyornithes are sister taxa, more closely related to each other than to
neornithines. It's especially surprising given the large amount of
characters supporting Carinatae and Gansus+Carinatae, though several are
unknown in hesperornithines or perhaps reversed due to their reduced wings.
860 steps- "Proornis" as an archaeopterygid
This is the classic position for "Proornis", though never supported by
synapomorphies. The reason it changes position with so few steps is merely
because so few (11) characters could be scored for it, and only two
confuciusornithid synapomorphies (manual ungual II much smaller than manual
ungual I; manual phalanx III-2 >150% of III-1 in length) were identifiable.
No archaeopterygid synapomorphies were identified.
860 steps- Elsornis closer to Euenantiornithes than Protopteryx
This was suggested by Chiappe et al. (2006). Supporting my basal placement
of Elsornis are the absent capital groove, short ulna, distally terminating
intermetacarpal space and unkeeled hypocleideum. It's possible these are
the result of neoflightlessess however. Supporting Chiappe et al.'s
arrangement are the laterally convex coracoid and concave central portion of
the humeral head. The addition of taxa like Iberomesornis which are more
derived but lack these characters makes the latter topology less likely
860 steps- Yanornis basal to Songlingornithidae + Ornithurae
This was proposed by Zhou and Zhang (2005), though it should be noted their
codings for most basal euornithines differ from those of Clark et al.
860 steps- Apsaravis closer to Carinatae than Hesperornithes
This was suggested in the original description of Apsaravis (Norell and
Clarke, 2001), but found to be slightly less parsimonious by Clarke and
Norell (2002). The change was due to correcting a few codings between
860 steps- Apatornis as an ichthyornithine
Only one more step is needed to support this position, originally suggested
by Marsh (1873), though there are no synapomorphies supporting it. As with
"Proornis", few steps are needed because Apatornis can be coded for so few
859 steps- Eocathayornis as an avisaurid
Zhou (2002) suggested both Eocathayornis and Sinornis (=Cathayornis) belong
in the same family. This was based on several characters, none of which are
included in this analysis. The more basal position shown here is due to the
less angled dorsal humeral condyle, apparently non-concave proximal
deltopectoral crest, elongate manus and manual phalanx III-2 present.
- From: David Marjanovic <email@example.com>