More on the 'cladistics' of sequences

[rjensen at SAINTMARYS.EDU]

Snip

Richard Jensen wrote:
> >Well, constructing a most parsimonious network and then rooting
> it by
> >noting the point where the outgroup(s) attach accomplishes the
> same thing,
>
John Grehan replied:
> I can see that rooting the tree after constructing a most parsimonious
> network may give the appearance of the same thing, but if the
> charactersare not limited to derived states through prior analysis
> of each character
> then my current conclusion is garbage in garbage out.

OK, then please explain to me how, in the absence knowledge of phylogenetic relationships, you determine which states are derived before conducting the analysis.

>
Jensen (cont'd):
> >but probably with less bias (i.e., the investigator is not making
> >arbitrary decisions about what constitute the ancestral states).

Grehan reply:
> If the decisions for what constitutes the ancestral states are
> verifiable then they are not arbitrary in the sense that the
> choices are not defensible.

Verfiable by what criteria?  And, whose definition of verifiable?

Grehan reply (cont'd):
>I content that if systematists know their group so
> poorly that they cannot
> make a documented argument to support the identification of potential
> synapomorphies before the analysis then I don't see how I could
> have any  confidence in the result.

I did not suggest that the systematist knew the group poorly.  It is essential that the systematist start with well-founded knowledge of group characters.  Obviously, once the analysis is conducted, it is necessary for the systematist to critically evaluate the result with respect to which characters provide support for the relationships.


Dick