[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]
Re: Ghosts of New Papers Past
----- Original Message -----
Sent: Wednesday, April 22, 2009 8:18 PM
>>>Hey, cool. BAND at last being done as science! At long last! :-) <<<
I'm afraid you spoke much, much to fast.
There I go, trying to make the best of my hopes public, and instead the
worst of my... realistic expectations get confirmed. No, surpassed,
I take everything back and call it a failure of peer-review... no, wait --
the failure lies with the editorial board for their complete failure to find
out where to even look for competent referees.
First, it's astounding that the publication date is 2009, as the
authors seem to have largely stopped reading the literature on homology
in the 1990's (or at least they feel that in typical BANDit style they
can quote papers from this time period and ignore or trivialize
Should have been automatic grounds for requiring a large revision before
Second, neither the authors nor the reviewer(s) have any knowledge of
phylogenetic analysis (as a third hypothesis, the reviewers could have
stopped reading in the first 10 pages, see below).
Why am I not surprised.
the cited literature they use already refers to it as the TWG matrix
(Or TWiG or TWIG.)
and defining Aves a priori as including Archaeopteryx
It's not bad they define it at all.
They also plan to not polarize their characters (e.g. not assume
that "0" is primitive and "1" is advanced...which is a perfectly
reasonable and interesting bit of data to publish)
Huh? PAUP* doesn't care about such silly assumptions. PAUP* makes _unrooted_
trees -- you see, at every change between 0 and 1, 0 is an autapomorphy in
one direction and 1 is an autapomorphy in the other direction --, then
arranges them in such a way that the outgroup lies on the outside, and then
calls that a rooted tree. The outgroup polarizes the characters, not any
people. That's what the outgroup is there for.
If the outgroup is all-1 instead of all-0, then 1 will be the plesiomorphic
state for each character. If the outgroup has states with different numbers
of different characters (the normal case), then which state is plesiomorphic
will differ between the characters. Which states you give which numbers is
arbitrary, it's just for your own purposes as a mnemonic or whatever you
Prior to their analysis, they discuss how they plan to address the
"problem" with assuming homology in characters where it "cannot be
established with certainty" BY PROCEEDING TO THROW OUT EVERY CHARACTER
WHOSE HOMOLOGY HAS EVER BEEN QUESTIONED BY A BANDIT.
I apologize for typing in caps, but please reread that and let it sink
And there I thought they'd at least superficially pretend to investigate
which homology assumptions are justifiable and which aren't! But no, they
just throw anything out that might look like work!
Essentially any important character of the manus, pes, or skull is
tossed out prior to the analysis because they don't want to burden the
analysis with "assumptions" of homology. They are committing the same
old (and repeatedly refuted) mistake of thinking that scoring
morphological similarity assumes homology, when in fact it TESTS
That's where the concepts of "primary homology" and "secondary homology" are
useful. You establish primary homology by the usual criteria (similarity of
shape, place, development, whatnot), then you code everything that's
primarily homologous as the same character, and then you run the analysis,
look at the resulting tree(s), and read from the tree which of the primary
homologies you found are secondary homologies (really derived from the same
thing in at least the MRCA) and which are homoplastic.
They claim to "test" their results later by including these characters,
but they score all the "questionable" characters with a ? (Which is
actually much worse and less honest, as it implies that the
morphological state is not known when in fact it is).
Looks again like they avoided all the hard work of finding out whether any
hypotheses on primary homology are justified or not. TSIB.
Worse, they then commit the exact crime they claim they are avoiding,
scoring "similarities" in tooth rooting in crocodilians and
Archaeopteryx (how do they know these are homologous???) and other
BANDit characters (ignoring the relevant literature on other
maniraptoran tooth socketing), thus "assuming homology" in their
Right when I thought it couldn't possibly get any stupider.
How stupid does one need to be in order to not notice the doublethink?
And yet _even_ worst, they claim they are testing the idea that birds
(and perhaps most maniraptorans) are not theropods by including
non-dinosaurian taxa, which would be fair enough, except they don't
actually sample the relevant taxa. They include Longisquama (with some
imaginary skull characters) but they don't bother to include any other
theropod except the ones that the TWG's tree was rooted on (Allosaurus
and Sinraptor). No basal theropods, no basal tetanurans. They prune
enough taxa away by the end to basically reduce non-maniraptoran
coelurosaurs to "tyrannosaurs" and all other theropods to "Allosaurus,
Sinraptor, and Ceratosaurus" and seem to not even grasp that this is a
Now that we know what doublethink they're capable of, that's suddenly not at
all surprising anymore. Doublethink is all you need to first (and
justifiably) complain that only including dinosaurs in a matrix doesn't test
whether maniraptorans are dinosaurs, and to then turn around and throw
practically all relevant dinosaurs out of the matrix.
The paper completely garbles phylogenetic methodology, but instead of
the classic BANDit tact of simply saying "cladistics sucks" they have
shifted their rhetoric (in a way more befitting of political
mudslinging than scientific discourse) to "we understand phylogenetics
better than you, which is why we are throwing out all of this data".
There's a word for that latter attitude: _egnorance_. I must stop here.
Google it or something.