Molecular taxonomy: on way out?

[John Grehan]

Derek Sites wrote:

> Mistaking a plesiomorphy for an apomorphy is a mistake not of the
> method but an error that can happen to a traditional morphologist as
> easily as it can happen in a molecular analysis (and I've seen plenty
> of poorly done 'traditional" morphological analyses as judged by well-
> done modern morphological analyses). 

I agree that mistaking a plesiomorphy for an apomorphy is not a mistake
of the method (which to my mind contradicts other assertions on the list
that it is the analytical algorithm that makes cladistics, not the
characters).

In traditional morphology the mistake comes from failure to identify a
character state also being present in the outgroup. At least this is
individually verifiable (testable) for each character. Correctly done,
the mistake does not happen.

When one uses an algorithm to make the correct choice the algorithm
cannot analyze the veracity of the individual character - it must take
the word of the systematist. If the systematist cannot limit characters
to derived states in the first place (as appears to be the case with
molecular studies) then the result is a relationship based on overall
similiarity of sequences some of which are 'chosen' by the alogorithm as
being informative relative to others in a process of cladistic mimicry.


Any phylogenetic analyses may
> sometimes make this mistake but this does not make the analysis
> phenetic.
It does if the characters are in reality a mix of primitive and derived
states that cannot realistically be distinguished simply by imposing an
outgroup on the date to magically transform some sequences into derived
states.

> 
> The point is that whatever state is found to be plesiomorphic,
> whether that is the truth or not, cannot be used to support groups in
> a phylogenetic analysis - such states are ignored. 

In a cladistic analysis they would not even be included in the analysis.
When characters are unordered one is basically declaring one's ignorance
and letting a computer algorithm somehow do magic.

Yes, ignoring them
> may be a mistake because they may actually be apomorphies, but,
> again, making this mistake does not make the analysis a phenetic
> analysis - because, of course, you must admit that traditional
> morphologists sometimes make this mistake too! (and doing so doesn't
> suddenly change their cladistic analysis to a phenetic one!)

It would if they used unordered characters so in effect saying that they
are clueless on the real state of the character and will let a computer
do the intelligence.

> 
> If both molecular and morphological methods sometimes produce
> erroneous character state polarizations, 

But if one restricts characters to those that are apomorphies the method
cannot produce erroneaous character state polarizations. The
polarizations are already given.

which we know can happen,
> then you cannot claim that one method is inferior (ie suffers an
> artifact) to the other for this reason. 

I can argue that failure to restrict the character set to apomorphic
states for the ingroup results in an analysis that is not cladistic -
only cladistic mimicry.

You may try to argue that one
> method is more likely to make this mistake more often than the other
> - but to make this argument you will need, I'm afraid, some evidence.

But I am not arguing that the problem is with the method! The problem is
with the failure to restrict the character set to apomorphic states in
the first place!

John Grehan