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

RE: pterosaur femora sprawl

David Marjanovic wrote:

> It's true there aren't any recognized flightless stem-bats, but what do you
> mean by "enormous range"? Archonta is dead, including the "flying primate"
> hypothesis, and the sister-group relationship of Chiroptera and Zooamata
> (Ferae + Perissodactyla) is supported by molecular data, including LINE
> insertion sites. Have I missed something?

(The following is a rather rushed response, but I hope I can get the point 
across.  Or at the very least, that it's not totally incoherent.)

In general, I think it's a good policy to approach molecular phylogenies with 
the same skepticism you would approach morphology-based phylogenies.  Both have 
their problems; but often they just happen to be different problems.  

One difference is that 'bad' morphological characters are often easier to 
detect than 'bad' molecular positions or sequences.  For example, with morpho 
analyses, we know which particular characters support a given clade, because 
they're there on the tree - and so 'bad' (plesiomorphic or homoplastic) 
characters are as plain as day.  A morphological analysis may reveal that bats 
(Chiroptera) and colugos (Dermoptera) are *not* sister taxa, in which case 
those characters associated with aerial locomotion would be deemed homoplastic, 
according to that particular tree.  Similarly, mesonychians were once united 
with whales because of shared dental characters between mesonychians and 
archaeocetes; further analyses (with more recently discovered fossil taxa) 
indicate that these characters are most likely associated with a similar diet 
rather than a uniquely shared ancestry.

Compared to morphological phylogenies, molecular phylogenies are a lot more 
opaque in this regard.  Unless you're willing to poke around the data at the 
level of nucleotides (for genes) or amino acids (for proteins), you'll never 
know *why* an analysis spits out a given topology.  It's just assumed that the 
software has been able to uncover the historical/evolutionary signal that's 
buried somewhere in the dataset.  But using mammals as an example, why do 
clades like Zooamata and Whippomorpha come up?  Why does the topology produced 
by the algorithm reflects the "true" (=historical) signal, and not just some 
bias (homoplasy)?  

Nucleotides and amino acids might look innocent enough, but the more you get to 
know them, the more you realize they can be very shifty characters.  
(Personally, I'm not sold on either Zooamata or Whippomorpha, although I do 
think whales arose within Artiodactyla -  just not from 'proto-hippoes').

In short, some people adopt an approach to molecular phylogenies that remind me 
of Bismarck's quote about laws and sausages (to retain respect for them, one 
must not watch them in the making.) 

>  From what I understand about LINEs, yes. LINEs are capable of inserting 
> twice in the exact same place,
>  but this is extremely improbable and has been found to be extremely rare.

I agree with Mike Habib here.  LINEs (long interspersed elements) are no silver 
bullet when it comes to reconstructing phylogenies.  LINEs (and SINEs) have 
shortcomings of their own.  Here's one treatment of the subject...

Hillis, D.H. (1999) SINEs of the perfect character.  Proc. Natl. Acad. Sci. USA 
96: 9979–9981.


Change the world with e-mail. Join the i’m Initiative from Microsoft.