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

Re: Morpho v molecular (was Re: Tinamous: living dinosaurs)

> Of course, somebody sit down with the matrix by Dyke &
> van Tuinen and actually test this speculation. And they
> should add a lot more fossils.
> The matrix by Livezey & Zusi has the same problems,
> especially a painful lack of fossils that cannot help but
> lead to long-branch attraction and other serious problems.
> It's just so big that fixing it would amount to _at least_
> an entire American PhD thesis... and thesis committees,
> AFAIK, tend not to like monomaniac projects.

I'm trying to take a look at it from the angle of "Falconiformes" specifically. 
But it'll mostly be the effect of partitioning, using the area around each 
appendicular joint as evolutionary units. It's possible because the matrix is 
so huge, meaning that you'll end up with >100 (often >200) characters per 
partition, which should be enough for a start.

But I have no idea yet if anything worthwhile (except in a strictly 
falsificationist sense) would result. In the end it may turn out that the 
problem cannot be tackled by quantitative analysis until a "missing link" 
fossil turns up. Higher-level Neoaves phylogeny is notorious for that, and 
mol-clock studies which claim a far older age nonwithstanding, the fossil 
record puts a lot of this radiation pretty damn close to the K-Pg boundary. 
(There is no bottleneck signal, but we know from Darwin's "finches" - which as 
volcanic-oceanic-island endemics *must* have had a bottleneck - that this 
signal is completely obliterated within a few Ma.)

The molecular data base for accipitrids and to a lesser degree falconids is 
quite great intraclade (if you add biogeography - of course this must be done 
quantitatively - it essentially resolves accipitrid phylogeny, only 
_"Accipiter"_ is undersampled). But it simply sucks interclade; molecular 
datasets do not tend to yield well-supported phylogenetic hypotheses 
*congruent/parsimonious with other data sources* here. So my main focus is 
whether partitioning can break the apparent convergence effect with owls, or 
whether the effect is simply too pan-bauplan.

The added benefit is that it is easier to include fragmentary fossils. If I 
find that some parts of the skeleton allow for a better/more robust/more 
congruent resolution of e.g. accipitrid pyhlogeny, I'll try to include as many 
accipitrid fossils (and presumed accipitrid fossils) as my datasets allow me 
to. It'll be interesting to take a quantitative look at all that 
Paleogene/early Neogene South American material. Hitherto it hasn't been done 
because it's usually just broken single bones, which add little beyond 
excessive computational load in whole-body analyses. With partitioned datasets, 
one can at least hope to determine whether the phylogenetic signal/noise ratio 
is >1; for example in the case of 
http://www.nrcresearchpress.com/doi/abs/10.1139/e01-057 it was evidently <1.

But this'll be stuff to do in 2013. Before that, there is also some molecular 
sampling to be done, to test the "harpiine eagles" and the "melieraxine hawks" 
more thoroughly. They may be clades - and would be interesting clades if they 
were - but it's not yet well resolved.

To the ways in which moldata can mislead, one would have to add a phenomenon I 
have seen times and again. It's similar to long-branch attraction. But while 
"classical" LBA is simply molecular convergence on the macro level, the other 
effect is convergence-caused misrooting, i.e. lower-level. I suspect the 
paleognaths in the Hackett et al. multigene phylogeny suffer from this: the 
topology is quite consensus if you unroot the paleognaths. But the "tinamous 
are secondarily volant" hypothesis suggested by their phylogeny stinks for 
obvious reasons (it violates everything we know about evolution of 
neoflightlessness in crown Theropoda). 

It may not be testable with molecular data because of the lack of bracketing 
taxa - no crown taxa between paleognaths and the Galloanseres-Neoaves split. 
Possibly, a massively diverse sample of Galloanseres and a whole lot of basal 
Neoaves would help (by more accurately determining ancestral charstates), but 
possibly not, and who's gonna test this anyway? So one can only surmise that 
the "secondary volant tinamous" are actually an artefact of "long branch 

(Truth to be told, I suspect a whole damn lot in the Hackett et al. phylogeny 
to suffer from this. Naive concatenation of multigene sequences is probably a 
Very Bad Thing, but this I'll definitely test. Accipitrids are good for this, 
because their extensive karyotype rearrangement is likely to have an 
hitchhiking/background selection effect above avian average.)