[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: Genes show Neoaves branching before K/Pg extinction

> > Polyphyly. Ratites could be paraphyletic only if the
> last common
> > ancestor of all paleognaths were a ratite, too -- a
> large flightless
> > bird with a keel-less sternum.

One major problem here: how do we know that ancestral-state reconstruction is 
correct? I have found (in sequence analyses) that rooting paleognaths correctly 
(and *knowing* it is correct) is almost impossible due to saturation. Sometimes 
it is clearly visible; this is the case when passerines turn up basal among 
neognaths. Otherwise, one can only note that the internal phylogeny of 
paleognaths (if you unroot them) is constant between loci, but if rooted it 
differs between loci.

Arguably RT might fix this, and the fact that they found 5 RT supporting 
_Struthio_ as basal among paleognaths is in line with this. But the tinamou-moa 
"clade" is unsupported by RT, particularly since the preceding node (the one 
excluding _Struthio_) has a large number of supporting RT.  (cf. Fig. 4 of the 

The fact that moa were not fully sampled for RT may be a reason (or THE 
reason), but I still wonder how many RT one would have expected for a 
moa-tinamou clade, given their prevalence on the other branches of the 
paleognath subtree, and taking into account branch length. Or, to put it 
another way, how unexpected zero-RT support for moa+tinamous is. Compare the 
Neoaves part of the tee, which looks as good as anything, and though RT are 
sparse here, their distribution is *not* conspicuously skewed (e.g. we don't 
have 5-RT support for Aequornithes).

The first phenomenon, BTW, might be the reason they only sampled one galliform. 
I found that getting cracids and phasianids to clade based on sequence data is 
quite tough already due to saturation. Often, the cracid will wind up in 
Anseriformes. Anhimids have a similar problem.

(To analyze sequences, I use PhyML latest version, with GTR and 10 rate 
categories, autoaligned in SeaView but with much manual correction as non-mt 
loci hardly align correctly over a wide taxo sample due to indels.)

In a nutshell, I have not found any locus to date (I checked all those for 
which a sufficient taxonomic sample for "higher landbirds" is available) that 
can be expected to root reliably once you get deep into the Mesozoic.

The paleognath problem, from what I have seen, can only be solved by:
1. reconstructing a consensus *internal* phylogeny (unrooted) from molecular 
2. morph analysis adding fossils, even if fragentary (partition dataset if 
3. Testing the consensus of 1. + 2. against various outgroup/locus 
combinations; you will find different rootpoints (= ancestral-state 
reconstructions), probably even for different outgroups *within the same locus*.
4. Analyze the pattern of rootpoints found in 4 and hope anything worthwhile 
turns up. E.g. clustering of rootpoints for various Galloanseres, with 
rootpoints for e.g. waders as outgroup outside this cluster, and for e.g. owls 
even further away.
(5. Wait for a konservat fossil of a Cenomanian-Turonian paleognath.)

The only shortcut I can think of: do a pure morph analysis using cranial 
characters *only* (well *maybe* throw in TMT and/or TBT) and including any and 
all K material available, no matter how fragmentary, as well as any and all 
early Pg material especially for paleognaths and Galloanseres.

It is easy to forget (one would think looking at the literature) that with the 
usual mutation rates, the null hypothesis is "around the K-Pg boundary, noise 
exceeds signal for most avian loci; my 100 Ma signal is essentially swamped by 
noise and any attempts to recover it will yield artefacts, perhaps more 
artefacts than signal." To presume you can simply go and root paleognaths with 
Galloanseres and expect it to be dependably correct is unrealistic, given that 
2 Galliformes + 2 Anseriformes do not clade correctly in many situations unless