Re: Underlying basis of classification (Was: Re Dinobirds)

[chris brochu <cbrochu@fmppr.fmnh.org>]

Philidor11@aol.com wrote:
> 

> 
> << In principle, *all* relevant characters are considered.  I understand that
> some analyses are more "selective" than others, but that's the fault of the
> analysis, not the method.  >>
> 
> My problem is the combination of 'all' and 'relevant'.  If characters are
> pre-judged as relevant then the analysis would be selective.  


OK, let me clarify - by "relevant," I mean "heritable and independent." 
I understand that independence is a real issue, but I do not believe in
tossing out characters because they seem unimportant.


Thinking of
> this as a computer program, is your model a program that would look at all
> characters, create a set of all possible trees consistent with the data, then
> select the 'best' among them according to certain criteria?  These criteria
> would presumably not consider one character to be a better indicator of
> relationship than another.

Yes.  And for morphological analyses, the criterion most commonly used
is parsimony - the tree that requires the smallest number of
transformations.



> 
> <<...the number of [shared] characters, by itself, is not an arbiter of
> closeness of relationship...
> Definition and diagnosis are entirely different operations, regardless of
> whether an ancestor is inferred or documented.  Dinosauria is the last common
> ancestor of ornithischians and saurischians and all of its descendents.
> Whether we actually have that last common ancestor in hand is irrelevant -
> after all, we don't have all of the descendents, either.  The diagnosis is
> the set of characters at that node in a given
> analysis - again, regardless of whether the state is actually known in all
> members; taxa for which the hip is unknown cannot be coded for the presence
> or absence of a perforate acetabulum, but other features support their
> placement in subgroups within Dinosauria.
> Diagnoses are not stable - and we have no expectation that they should
> be.  Any new discovery will change them.>>
> 
> I wasn't arguing for a cumulative count of shared characters as decisive, but
> I am curious about how a much less than complete connection can be diagnosed.
>  You mention other features, but they can't be directly associated with the
> ornithiscian hip, which the non-bird-hipped ornithiscian does not have.
> Aren't you therefore creating a second diagnosis of ornithiscia, arguably
> better than the one based on the hip, because it includes at least a
> substantial number of the animals you want to include, but who are left out
> by the hip-based definition?

No.  I can explain better with an example:

Suppose we find a partial skull in the field.  It shows clear evidence
of horn cores, an expanded frill along the posterior margin of the
skull, and a rostral bone articulating with the predentary.  Any
analysis in the world would place this fossil within Ceratopsia.  So
even though we don't have the pelvis, we know it belongs to a group
that, on the basis of evidence from other specimens, is deeply nested
within Ornithischia.

One of the most common misconceptions out there about phylogenetics is
that incomplete fossils will always be useless or, worse, positively
misleading.  They can be, but they more commonly cause local losses of
resolution without adversely affecting the rest of the tree.



> Also, how does the incomplete record of descendants ('after all, we don't
> have all the descendants') justify the inability to document an ancestor
> ('[w]hether we actually have that last common ancestor in hand iis
> irelevant')? 

By itself, it doesn't.  We justify the inability to document an ancestor
because, frankly, we can almost never do it.  The only positive evidence
for ancestral status of a species would be the discovery that some
members are actually closer to another species than to other members of
its own species, and that requires a massive sample and is almost never
reflected in morphology.  We can make hypotheses of ancestral status on
the basis of negative evidence - "Fossil A does not occur above fossil
B, and does not have autapomorphies, and is the sister taxon to fossil
B, so fossil A might be the ancestor of fossil B."  We could always
later discover fossil B below fossil A, or autapomorphies in fossil A,
that would falsify our hypothesis.




 This seems instead to weaken the case for the ancestor because,
> as you point out, the better the fossil record the better the analysis.


More accurately, the better the taxon sampling, the better the analysis
- whether the ingroup taxa are living or extinct should make no
difference.



> You identify my core problem when you assert that '[d]efinition and diagnosis
> are entirely different operations', because, absent an ancestor, the
> definition is created, defined by the diagnosis of the significant
> characters. 


No, it isn't.  As an example, have a look at the first issue of
Systematic Biology for this year, in which several taxa within Mammalia
are defined, all on the basis of trees derived from molecular data. 
These groups conflict with morphological analyses - so there aren't any
characters "defining" them.

Moreover, if you follow the literature for any specific group - crocs,
dinosaurs, mammals, whatever - you will often find the overall set of
relationships remaining more or less stable, but the diagnoses changing
wildly.  The presence of feathers used to diagnose Avialae.  Now we know
feathers diagnose a more inclusive clade within Theropoda, since they're
present - in one form or another - in several nonavialean coelurosaurs. 
But the definition of Avialae is the same, and the shape of the tree in
that part of Theropoda is no different.  This is because the definition
of Avialae has nothing to do whatsoever with the characters diagnosing
the group - it would include birds and those taxa closer to them than to
dromaeosaurids regardless of the characters at the root of Avialae.



 Further, the identification of the significant characters also
> seems subjective if proposed by the analyst, as discussed above.
> 
> <<Proven?  Can't be done.  But as scientists, we're not in the business of
> proving anything - its disproof we're after.>>
> 
> You mean that hypotheses can be refuted only by proving the negative?  That's
> difficult to do, particularly if one of the problems is absence of data.

Not what I mean.  A tree can be overturned with new discoveries.  The
tree is not a "negative," since it is initially based on some sort of
hard evidence.



> 
> <<As far as phylogenetics is concerned, only two factors are important - that
> the characters be heritable and independent.  If they're passed on from
> parent to offspring, they can preserve a phylogenetic signal, regardless of
> subjective assumptions of "importance.">>
> 
> Aren't you dealing with characters which are heritable and are independent,
> but which also might be coincidence in unrelated lineages? 

Sure.  But unless you know the phylogeny a priori, how would you know?



 This seems to
> argue that a hypothesis is correct because it is possible?  

No - only that a hypothesis is favored by the data set at present, and
will stand or fall with later work.



I was asking if
> one way to test for whether a shared character is significant might not be
> whether the character helped the animal survive?

And you would test this for a fossil by......?

In fact, some have argued for NOT using these characters, because they
are likelier to be favored by selection and to reoccur in different
lineages.  I don't accept that line of reasoning, but it's out there.




  In fact, I was wondering
> whether I should assume that a character must be part of the survival
> strategy of an animal in order to use the character to identify a kinship
> with another animal.


I don't think so, because (a) we can usually not know this and (b)
non-selective characters also preserve a phylogenetic signal, and we
don't want to ignore something that might help us work out
relationships.



> 
> <<Detailed morphological analyses can highlight differences questioning
> assumptions of homology.  Otherwise, we can't know two identical features are
> nonhomologous without a phylogeny.>>
> 
> I'm still working on this, but when you say 'we can't know two identical
> features are nonhomologous without a phylogeny'  are you saying that only a
> secure theory can prevent features from being considered homologous, in the
> absence of decisive morphological analyses?  


More or less.  Saying "two characters are not homologous" makes an
explicit statement that the last common ancestor (inferred or otherwise)
of the taxa in question did not have that feature - and that assumes
that you know the phylogeny.

hope this helps.  Let me know if I continue to misunderstand.

chris


-- 
----------------------
Christopher A. Brochu
Department of Geology
Field Museum of Natural History
Roosevelt Road at Lake Shore Drive
Chicago, IL 60605

voice: 312-665-7633  (NEW)
fax: 312-665-7641 (NEW)
electronic:  cbrochu@fmppr.fmnh.org