[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]
Thoughts on the biggest morphological bird analysis
Livezey and Zusi, 2007. Higher-order phylogeny of modern birds (Theropoda,
Aves: Neornithes) based on comparative anatomy. II. Analysis and discussion.
Zoological Journal of the Linnean Society. 149 (1), 1-95.
Well, Livezey and Zusi's massive effort is finally published, and I do mean
massive. 2954 characters, from all areas of morphology, make this the
largest bird morphological analysis by over an order of magnitude. In
addition, nearly every non-passeriform family of living bird is included, as
well as a huge amount of outgroups and a few subfossil taxa. The paper is
freely available here- http://www.blackwell-synergy.com/toc/zoj/149/1 , and
I strongly recommend reading it. Their Consistancy Index of 0.2432 is very
reassuring, as it indicates unplanned results (as well it should, with so
many characters). Ordering seems to have been done properly, and taxon
choice seems commendable. This is THE way to do a morphological analysis.
I'm extremely impressed.
Livezey and Zusi used 35 outgroups, most of which are theropods. It's at
this point I wish I had the companion paper, which describes the characters
used, because I don't know if the authors included characters traditionally
diagnostic of nodes like Coelurosauria or Theropoda, or if they analyzed
non-avians in an avian matrix. The outgroup phylogeny is fairly standard,
with the exception of Phytodinosauria (58% bootstrap), Troodontidae as
Avialae (<50%), Alvarezsauridae as close to birds as Archaeopteryx (98%),
Rahonavis in Ornithothoraces (69%), Hesperornithes sister to Ichthyornis
(62%) and Lithornis outside Aves (91%). The position of troodontids and
hesperornithines is quite possible in my opinion, but clearly oddities here
"reflect missing key taxa and poor preservation of those coded", as noted by
the authors. A number of taxa were pruned and merged (pg. 9), but surely
Shuvuuia should have been merged into Alvarezsauridae, Segnosaurus and
Erlikosaurus into Therizinosauroidea, and Gobipteryx into Enantiornithes
(instead these four genera were pruned). For those fans of taxonomy, we get
a lot of new names. Troodontoidea, Dromaeosauroidea and Alvarezsauroidea
are all used where their familial counterparts should have been. Also named
are Rahonaviformes, Rahonavidae, Apsaraviformes and Apsaravidae. I'm not
sure of the use of these, as Rahonavis is probably a dromaeosaurid,
unenlagiid or jeholornithid, while Apsaravis is quite possibly an ambiortid.
Another new name is Palaeoaves, which seems to be a paraphyletic grade of
taxa from Archaeopteryx to Apsaravis. Why Livezey and Zusi use so many
paraphyletic groups outside of Aves I'm not sure (e.g. Archaeopteryx and
Confuciusornis are joined in Archaeornithes; Apsaravis and Rahonavis in
Euenantiornithes- gag), but I don't approve. In any case, non-avians were
not the point of this analysis, so I digress.
The avian phylogeny is very well structured in the strict consensus tree,
but many nodes (especially deeper ones) have terrible bootstrap support.
Bootstraps have their problems, but with most every avian known from
complete specimens in this analysis, they should be an accurate report of
rearrangment liklihood. I find it funny that Livezey and Zusi say things
like "The present data set (Livezey & Zusi, 2006) supports the rejection of
this novel proposal involving the grebes and flamingos, and suggests that
the taxonomic proposal for the couplet by Sangster (2005) is premature",
when they name taxa like Terrestrornithes which has a bootstrap of <50% and
a Bremer support of 1! Indeed, their taxonomic scheme overflows with
redundant taxa, several of which are newly named- Casuariimorphae,
Struthionimorphae, Hemiprocni, Trogonomorphae/Trogones, Pico-clamatores. I
agree with David regarding their taxonomic disaster. If we hold Livezey and
Zusi's analysis to the same standards as huge molecular analyses in regards
to support (say 95% bootstrap liklihood), nearly all the major nodes
| | |--Struthionidae
| | `--Rheidae
| | |--Anhimidae
| | `--Anseres
| | |--Anseranatidae
| | `--Anatidae
| | |--Gaviidae
| | `--Podicepididae
| `--Procellariidae (incl. Diomedeidae)
| | |--Jacanidae
| | `--Rostratulidae
| | |--Glareolinae
| | `--Cursoriinae
| `--Columbidae (incl. Raphidae)
What this says to me is that the relationships of major neoavian taxa may be
forever beyond morphological resolution if only recent taxa are included.
Interestingly, there are only a few areas in which this disagrees with the
largest current molecular analysis of Neoaves (Ericson et al., 2006; 5007
characters) if the latter is evaluated at 95% (Bayesian posterior
probability this time). Ericson et al. had similar taxon sampling to
Livezey and Zusi, except they included less Charadriiformes and outgroups.
- Grebes are related to loons instead of flamingos. It takes 146 more steps
to get Mirandornithes in Livezey and Zusi's matrix, even though I assume
Mayr's (2004) characters were included. The present authors foresaw my
opinion- "it is predictable that strong confirmation of a sister-group
relationship between the Gaviiformes and Podicipediformes herein will
engender concerns of artefactual pairing by convergence". I agree with many
of Livezey and Zusi's points, that many previous criticisms of the
loon-grebe clade (newly named Gaviomorphae by the authors) were flawed, and
that the comparatively poor morphological evidence for Mirandornithes would
surely lead to its rejection if morphology were all we had. Yet Ericson et
al.'s support for Mirandornithes is impressive, as is that for the five
additional clades separating grebes and loons in his topology. Furthermore,
comparison with mammalian phylogenies suggest that convergent taxa group
together in morphological analyses, sometimes with large support indices.
In Asher et al. (2003), two of the best supported clades are bats plus
flying lemurs, and pangolins plus anteaters. Much as I'll continue to
support these groups being convergent, I'll support grebes and loons being
- Albatrosses are derived procellariids instead of diving petrels being
derived procellariids and albatrosses being the basal procellariiformes.
- Tropicbirds are pelecaniformes instead of metornithines.
- The shoebill is a pelecaniform sister group instead of a hammerkop
relative. It only takes 30 steps to move the showbill inside
Pelecaniformes, presumedly by pelicans. Placing the hammerkop by or inside
of Pelecaniformes only takes 23 steps. Neither of these is precisely
congruent with Ericson et al.'s topology, but it does seem to indicate the
taxa are fairly labile.
- Pelecans are with traditional pelecaniforms instead of by the shoebill and
- Seriamas are sister to bustards instead of parrots+falcons+passeriforms.
- Buttonqail are outside Charadriiformes instead of nested inside it. Here
the large amount of Charadriiformes included by Livezey and Zusi (21 vs. 3
in Ericson et al.) may make you think the position of buttonquail is an
artifact, but more extensive molecular analyses of charadriiformes (e.g.
Paton and Baker, 2006; 12902 characters) agree they are the sister group of
Lari, with 100% posterior probabilities in almost every node.
- Falcons are related to other diurnal predatory birds instead of
- The hoatzin is sister to cuculids and musophagids instead of a
- Passeriformes are in a trichotomy with piciforms and coraciiforms instead
of seriamas, falcons and parrots.
- Cuckoo-rollers are coraciiformes instead of more basal.
Even many of the larger clades which aren't well supported in Livezey and
Zusi's work correspond to clades in Ericson et al. though.
Though the former did not recover Metaves, it recovered Cypselomorphae at
75%. Within that clade, Apodiformes is well supported at 100%. It only
takes 31 more steps to get Daedalornithes. Other supported metornithine
clades are Columbiformes (100%) and Eurypygae (77%). Thus Metornithes
really does seem to be the Afrotheria of birds, as afrotherian clades
(Tenrecoidea, Paenungulata, Tethytheria) are recovered morphologically
though there's no known morphological support for Afrotheria as a whole.
As noted above, Charadriiformes is recovered (97%; albeit without
Turnicidae). Yet, turnicids are a possible charadriiform sister group in
Livezey and Zusi's tree, with only ralliforms more closely related (at <50%
certainty). Within Charadriiformes, recurvirostrids and haematopodids are
sister taxa at 64%, with the other Charadrii not separated by nodes of over
50%. Rostratulids and jacanids are strongly supported sister taxa (99%)
with the other Scolopaci all placed as the most basal members of the order.
So this is probably a rooting problem. Lari (99%), Laroidea (100%), and the
larid+sternid+rhynchopid clade (91%) are all well supported, with glareolids
not separated by nodes of over 50%.
Grues (100%) and Gruoidea (100%) are both well supported, as is Ralliformes
(100%) and are located nearby on the cladogram with otidids. Cuculids and
musophagids are weakly joined (66%) and are potential sister taxa in Ericson
et al.. Besides the two metornithine families, Livezey and Zusi's Natatores
(<50%) corresponds to a clade in Ericson et al.. Procellariiformes is well
supported (100%), with the Procellariiformes+Sphenisciformes clade (88%) and
the clade joining this to Gaviiformes (100%) are not contradicted by the
molecular data. Phalacrocoracoidea (100%) and Sulida (87%) are supported.
Livezey and Zusi's Ciconiimorphae (78%) is mostly represented by the
molecular data, though missing Pelecanidae and Balaenicipitidae, and with
Ciconiidae itself only a possible member. But balaenicipitids are the most
basal members of the ciconiimorph's sister taxon in Livezey and Zusi, so
they're not far away.
Besides the metavian Cypselomorphae, Opisthocomidae and Columbiformes, and
the gruiform-related Cuculiformes, Livezey and Zusi's Dendrornithes (58%) is
represented in the molecular data. Ericson et al.'s
Cariamidae+Falconidae+Psittaciformes+Passeriformes clade is another one
which receives no morphological support, though its components do
(Falconidae 99%, Psittaciformes 100%, Passeriformes 100%).
Strigiformes (100%), Sagittariidae+Accipitridae+Pandionidae (89%), and
Accipitridae+Pandionidae (100%) are all recovered, and the molecular data
leaves the possibility of Raptores (72%) and Accipitriformes (100%) open.
Thus the entire topology of raptorial birds could correspond, excepting
Falconidae's absence. Livezey and Zusi's Trogones+Pico-Clamatores clade
(74%) is represented in Ericson et al. except for the absence of
Passeriformes in the latter. Coliidae and Trogonidae are joined in Livezey
and Zusi's tree (78%), and are potential sister taxa in Ericson et al.'s.
Bucerotes (98%) and Upupides (100%) are found in both. A Halcyones+Coracii
clade (76%; though minus leptosomids), Coraciides (96%; though minus
leptosomids), and Halcyones (97%) are found in both. Coracii (63%) is not
contradicted by the molecular data. The entire topology of Piciformes (88%)
is shared by the two studies- Galbulae (100%), Pici (89%), Capitonoidea
(<50%), and Picoidea (51%).
I have to agree with David regarding Livezey and Zusi's fear of missing
data. There's no reason Dinornithiformes and Aepyornthiformes should be
excluded when we have complete skeletons. Taxa don't cluster together due
to missing data, and inserting them a posteriori won't let their unique
character combinations improve the tree. I think adding fossils to this
tree will be extremely interesting, especially if the molecular data is
added and analyzed simultaneously. Fossils are the morphologist's only hope
at combating such detailed convergence.