Message 2004-10-0059: Apomorphy-based clades; was Re: Panstems

Tue, 14 Sep 2004 05:44:42 -0700

[Previous by date - Re: Panstems]
[Next by date - Re: Fwd: Re: Panstems]
[Previous by subject - Anti-Phylogenetic Comments in The Botanical Review (part 2)]
[Next by subject - Apomorphy-based definitions]

Date: Tue, 14 Sep 2004 05:44:42 -0700
From: [unknown]
Subject: Apomorphy-based clades; was Re: Panstems

Kevin de Queiroz wrote-

> In addition, potential homoplasy in the character is NOT a good rea=
son for
avoiding an apomorphy-based
> definition (and in this case, Michel seems to be disregarding comme=
that were made at the Paris
> meeting).  If other taxa have evolved the character convegently, th=
taxa can be excluded by specifying
> that the condition has to be homologous with that in mammals (or
Varanops).  If instead the character arose > earlier, then we would s=
have to conclude that some other early amniotes that have not traditi=
> been considered synapsids are, in fact, synapsids (changes in hypot=
composition happen all the
> time).

I was unable to attend the Paris meeting and have not read all of the
literature debating the use of apomorphy-based clades, but perhaps yo=
u could
enlighten me as to why these arguments are invalid-

1. A philosophical argument.
Phylocode is an extension of phylogenetic taxonomy, the practice of u=
phylogeny alone to determine taxonomy.  Thus, taxonomy should only ch=
when phylogeny changes.  If this isn't so, phylogenetic taxonomy is c=
not purely phylogenetic, and is instead tainted by anatomy (used in t=
sense of all physical structure, including genetics) and/or behavior.=
latter two form the major Linnaean justifications for taxonomy.

2. Continuity again.
Avialae was traditionally defined as (Ornithurae <- Deinonychosauria)=
Perhaps in strict Phylocode rules it would be (Vultur gryphus <- Dein=
antirrhopus).  It was later also used for the clade (Archaeopteryx + =
a node-based definition that is basically equivalent to the original
definition in known content.  Two decades later, it was redefined as =
apomorphy-based clade- (feathered wings homologous with Vultur gryphu=
s and
used for powered flight).  This seemed like an equivalent clade at th=
e time,
since no flying deinonychosaurs were known, and Archaeopteryx was tho=
ught by
most to have flown.  I assume much thought went into this definition,=
there was a good half page written defending it, in a paper whose pur=
was to define several bird-related clades.

Then in 2002 Microraptor was discovered to have as many characters
associated with flight as Archaeopteryx, and then some.  Yet Microrap=
tor has
consistantly fallen out in basal Deinonychosauria (Senter et al., 200=
4; Xu
et al., 2000; Theropod Working Group 2002-2004).  So now deinonychosa=
seem likely to be basally volant, with flight developing at the
Deinonychosauria + birds node (Eumaniraptora) at least.  And under th=
apomorphy-based definition, this means deinonychosaurs are avialans.

You would probably say "we would simply have to conclude that some ot=
theropods that have not traditionally been considered avialans are, i=
n fact,
avialans (changes in hypothesized composition happen all the time)." =
because changes in hypothetical composition occur does not mean we sh=
make them any more likely to, if we can help it.  Forbidding apomorph=
definitions eliminates one source of potential disruption to continui=
Continuity is a major concern of many advocates of Phylocode, as well=
those who aren't as quick to embrace it.  And this is quite the chang=
e in
continuity, as Avialae would now include the only group that was ever
specifically excluded from it.

Importantly, this kind of change in hypothetical composition is NOT d=
ue to
phylogeny, but instead due to anatomy and behavior.  If we find
deinonychosaurs are more closely related to birds than to Archaeopter=
yx (as
in Paul, 2002), forcing them to become avialans in the node-based sen=
taxonomy is changing to reflect phylogeny.  No new morphological info=
is needed for this change, one could recognize it from the topology a=
But if we find out basal deinonychosaurs flew, forcing them to become
avialans, taxonomy is changing to reflect morphology.  The topology i=
s still
identical to when before Microraptor was known.  Quite the regression=
Linnaean times, when taxa were discovered to be too disparate from th=
relatives, and thus placed in another family or order.

3. Good ol' incompleteness.
Another problem is that fossil taxa never preserve everything about t=
anatomy and behavior.  Using "powered flight" to define a clade is a
particularily egregious example, as the flight capabilities of fossil=
are virtually never certain, and are especially ambiguous when first
developing.  We haven't had over a century of debate regarding
Archaeopteryx's flight abilities for nothing.  Even if you accept
Archaeopteryx flew, the next furthest taxon from the crown is
Scansoriopterygidae (according to Zhang et al., 2002 and Czerkas and =
2002).  But this is only known from two juveniles, making flight
capabilities of adults unknown.  How are we to apply Avialae unambigu=
This can be mitigated by using those structures that preserve frequen=
but even that is problematic.  The apomorphy-based Ornithurae is base=
d on
pygostyle structure, but the latter is unknown in possible basal
ornithurines Patagopteryx and Vorona.  One of the supposed advantages=
Phylocode is that the definition of clades will stop the common Linna=
debates as to the extent of groups of organisms (e.g. is Archaeoptery=
x a
member of Aves or not?).  Apomorphy-based definitions lead to the sam=
subjective debates, though now in the form of character distributions
(ACCTRAN vs. DELTRAN or something in between?).

4. Confuciusornis and Carinatae.
Indeed, such debates take up their traditional form due to yet anothe=
problem of apomorphy-based definitions- continuously variable charact=
being used as apomorphies.  A good example is Carinatae, recently red=
as (keeled sternum homologous with Vultur gryphus).  The authors of t=
definition used Chiappe et al.'s wording to defend the structure in s=
Confuciusornis as a ridge and not a keel, thus confuciusornithids are
outside Carinatae.  They left it to future authors to decide where th=
precise boundary is between ridge and keel.  That of Protopteryx seem=
s a bit
more developed than Confuciusornis, but less than more derived taxa l=
Concornis and Ichthyornis.  So is Protopteryx a carinate?  Any charac=
will end up showing continuous variation once we get enough specimens=
, and
objectivity is not so easy to provide.  Confuciusornis sanctus shows =
us the
population varies in respect to carina height, with some having no ca=
and others having one.  We can hardly measure carina height for the v=
majority of relevent specimens, as they are crushed into slabs.  This=
to ambiguity over which taxa are different enough to exclude from the=
exactly as in Linnaean times.

5. Muller and Diapsida.
Finally, there's the homology argument.  It's not easy at all to dete=
the homology of most features, due to the widespread homoplasy in nat=
As an example, it was recently proposed at the Paris meeting that Dia=
be redefined as (1st reptile with Caiman crocodilus' two temporal
arches/fenestra).  Now, for the last century, Diapsida has included
basically all amniotes not on the lines to turtles or mammals.  The n=
eed for
monophyly has necessitated the inclusion of birds, euryapsids, and ma=
turtles as well (though not their traditional relatives).  The tradit=
hallmark of diapsids has been their two temporal fenestrae, which are
usually thought to be homologous across the clade.  But recently Mull=
(2003) suggested that the lower temporal arch of recent diapsids was
convergently developed after more basal diapsids lost it.  This was
supported by anatomical evidence and parsimony in character state
distribution.  Indeed, parsimony is ambiguous as to whether the lower=
of archosaurs and lepidosaurs is even homologous.  Using DELTRAN,
apomorphy-based Diapsida would include only archosauriformes and
rhynchosaurs.  Using ACCTRAN, it would include only Sauria.  Neither
optimization favors a content anywhere close to what Diapsida has bee=
n for a
century.  And it's ambiguous which optimization we should use!  Homol=
ogy is
nearly always decided on such simple concepts as parsimony, where one=
branch can change the homology assessment.  Do we let lepidosaurs alt=
between being diapsids or not based on the number and sequence of mor=
e basal
taxa with a lower temporal arch?  I think the concensus would be no.

People defending apomorphy-based clades generally say these problems-
1. Can be solved by better definitions or redefining clades.
2. Are not unique to apomorphy-based clades, since there will always =
be taxa
of uncertain taxonomic placement.
However, I have yet to see justification for why we should bother to =
clades that could require redefinition.  Based on the results of Avia=
lae and
Diapsida, even the most careful and planned apomorphy-based definitio=
ns seem
incapable of preserving traditional clade content when new data is
discovered.  Would it not be better to limit ourselves to clades that=
have stable definitions?
Similarly, I have yet to see justification for why we should retain a=
of taxonomic ambiguity when it's very easy to eliminate that source. =
only taxa that would have ambiguous taxonomic placement would be thos=
e with
ambiguous phylogenetic placement, hence phylogenetic taxonomy.
It's true that apomorphy-based clades are superior to Linnaean method=
s in
the ways I compared the two above, since at least DELTRAN vs. ACCTRAN=
 is an
objective distinction, and the kind of anatomical/behavioral divergen=
needed to separate specific clades is made clear.  But neither proble=
m is
faced by stem- or node-based clades, so why bother with apomorphy-bas=
definitions at all?
It's basically what Laurin was saying before, we can do better, and w=

Mickey Mortimer
Undergraduate, Earth and Space Sciences
University of Washington
The Theropod Database -


Feedback to <> is welcome!