Re: Combined answer 1: cladistics

["David Marjanovic" <david.marjanovic@gmx.at>]

> >Obviously it doesn't have a monopoly on accuracy -- or if it does, we
can't
> >find that out, because to do so we'd have to compare cladograms to The
True
> >Tree. But... what other scientific methods are there for phylogenetic
> >reconstruction? I know none.
>
> Inferences on phylogeny derived from any given area, be it genetics,
> embryology, or what have you, are if formulated in such a way that they
make
> a specific prediction which can be falsified, as valid as any PAUP
generated
> cladogram.

As Mickey has mentioned, "cladistics" does not automatically equal
"morphological analysis using parsimony". Molecular evidence is _always_
formulated as a cladistic data matrix, even when it's phenetics (except
DNA-DNA hybridization and immunological comparisons), be the program PAUP*,
PHYLIP, NONA, Hennig86 or whatever. When you read "Genetic evidence that
hippos and whales are sistergroups", you know it's cladistics (be it
parsimony, maximum likelihood or Bayesian). I haven't seen explicit
phylogenetic hypotheses based on embryology that were not cladistic
(Feduccia et al. don't produce phylogenetic hypotheses, as discussed).

Regarding subjectivity... oh yes, cladistics can't completely eliminate
subjectivity. But it does reduce and confine it -- it forces one to make it
explicit. To use a hypothetical example, if you would weight a character,
you'd have to write that you weighted it, how much and why. The "how much"
is important to keep the analysis _reproducible_: the same data set should
give the same tree(s) when analyzed under the same criteria, no matter who
repeats it with which program. In addition, each cladogram is an _explicit_
hypothesis that says "I explain my data matrix _most parsimoniously_"; this
can be tested by repeating the analysis as described.
        Back to the beginning... it has become (and should always have been)
commonplace to describe in detail how one has -- subjectively -- coded one's
characters, and why. When I write "I coded *Apsaravis* as lacking
premaxillary teeth because the tip of its dentaries is toothless", then
people can point out that *Sapeornis* has toothless dentaries but toothed
premaxillae and maxillae, falsifying my hypothesis that these characters
were correlated and should be treated as one. If I use a program that cannot
deal with inapplicable character states (like PAUP* or Hennig86 as opposed
to NONA), I have to explain how I avoided inapplicable codings.