Positivism vs Realism

[Richard Zander]

Thomas Pape wrote:
>
> At 13.58 1997-12-12 -0800, Zander wrote:
>
> >   Cladistic
> >convergene is homoplasy that occurs for [?far] enough away that convergence
> >cannot be interpreted wrongly as due to shared ancestry. Take a true
> >tree ((a b) b), where character b is an advanced trait, and is evolved
> >twice patristically near. "Parsimony analysis" necessarily interprets
> >the true tree as ((b b)a). Patristically nearby homoplasy is verboten in
> >maximum synapomorphy analysis.
>
> Well, did you say "terribly simplistic evolutionary theory" ?? I would say
> that your example is just too 'terribly simplistic' being based on only two
> states. Your "patristically nearby homoplasy" is by no means "verboten" in
> max parsimony analysis. If other evidence is in favour of the ((a b) b)
> tree, this is what we accept. And if proper outgroup comparison will
> indicate that state b is apomorphic at the level of your three taxon clade,
> state b will be considered to either have evolved twice -- or to have
> evolved once and been reduced (or transformed) once.
>

Well, hold on now. I gave a simple example and you are somehow making
this a simple theory. Cladistics weighs numbers of state changes and
decides in favor of the majority. You cite additional evidence often
being available as being able to support the convergence I suggested.
Great! I would like to know how much support is enough support: ((acd
bcd} b)? ((acdef bcdef) b)? Of course "it depends," for instance on how
much cdef might be interpreted as a suite of characters evolving as one
in response to a particular habitat, in which case you have only one
character support...etc. Suddenly we are in realm of judgement calls and
everything else that numerical methods are supposed to get away from.
Okay, we can't avoid judgement. The question is, to what extent can we
eliminate it. I opine that because cladistics is open-ended as far of
confidence limits go, maximum synapomorphy is just that, not maximum
parsimony, in cases that are in any way arguable. We all accept
"accepted" trees; this is why we do not include dog data in our analysis
of fish or daisies. Yes, cladistics can help decide certain puzzling
problems, but exactly when can it do so and when can it not do so?
Surely you don't believe that cladistics can solve all phylogenetic
problems for which there is data?

> Indeed, this allows the possibility of BOTH convergence AND common ancestry.
> But how should ANY other analytic method be able to distinguish between
> these two possibilities?
>

How indeed. Perhaps they can't. To what extent then are our cladograms
and statistically derived trees valid?

--

*******************************************************
Richard H. Zander, Buffalo Museum of Science
1020 Humboldt Pkwy, Buffalo, NY 14211 USA bryo at commtech.net
*******************************************************