[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: 2 questions about characters (long!)

Charles Roustan <charlesroustan@softhome.net> wrote:

> Are characters with "intermediate apomorphies" the same as multistate
> characters?  and when one code these, what is the guide one is supposed to
> follow to decide to code it as multistate or as your regular 2-faced
> character after all, any quantitative or qualitative, or whatever
> can be transmogrified into a 2-state one with better or worse results, or
> I wrong? what's the good in these multistate characters? I can think of a
> couple of examples but perhaps you could enlighten me with another?

I'm not entirely sure what an "intermediate apomorphy" is supposed to be,
but it sounds like a state in a multistate character to me. There really
isn't much of a "guide" in how to code characters. A lot depends on one's
philosophical outlook on what characters are and how we should use them in
analysis. Amusingly enough, this is a topic of some of my PhD research...
for right now, the best I can tell you from my limited experience, is that
no one knows for sure. A great way for you to explore this is to get hold of
a program (such as PAUP) that allows you to use these options, and fiddle
with real data. Look at what happens to the tree when you change character
codings. Also look at papers and real data: compare the character
descriptions and the distribution of states in the matrix with the
morphology observed in the specimens. Examine how other people parsed
morphology into characters, and decide how you feel about their choices. My
personal belief is that you need to look at how much information is being
captured by the coding, but others would argue that you have to try and
minimize assumptions about how the characters evolved when you code them.

A reasonably good primer, which I believe is out of print, is the Compleat
Cladist... it is more helpful than a lot of textbooks, and will get you into
the guts of parsimony methods with relatively few headaches.

> >> If I didnt know
> >> something had something, I "assumed" it was a no. Next
> >> time I'll make sure the info is more accurate
> To which Mickey Mortimer answered-
> >Needless to say, that's a bad practice.
> Now, I agree, of course, is this the same as saying that everytime one
> doesn't know a particular character state, one should use a question mark,

    Absence of a particular feature, even if you feel it is the ancestral
state, could very end up being a reversal that unites a group of organisms.
The classic example: presence of limbs, in the context of a phylogeny of
vertebrates, is a derived state, yet absence of limbs unites snakes and
several other groups of vertebrates. Therefore, coding taxa in which you do
not know the state as "absent" misrepresents the data available; because
that absence CAN become a synapomorphy supporting alternate topologies.
That's the really really simple version, anyway.

> I know (I'm quoting again here but I don't remember who said it) that any
> cladistic analysis is just as good as its choice of characters and
> apparently the choice of characters is a function of how much you actually
> know about the groups you are fiddling with (how much you, the analyst,
> knows and also how much the entire scientific community knows about the
> given groups).

    Character choice is a tricky thing. Some workers (implicitly or
explicitly) choose characters based on their conception of how "valid" or
important or informative the character is, while others try to amass every
possible observation they can make and code them. In both cases, familiarity
with the group is important; in the former case, it is vital (you must
develop "special knowledge" before you can use it... that was sarcastic, by
the way), in the latter case it is helpful for finding new variation to
code. There is a paper by Poe and Weins in which they showed that an
alarmingly small number of studies gave explicit criteria for character
recognition. My impression has always been that there's often a bit of
voodoo involved...

> To end this, I would like to ask you for some references that help me get
> little better. I've tried many books, old and newish with the same
> maybe I just need time?

    I hate to say it, but time will probably help the most... time and
reading. There are a reasonable number of papers on character choice and
character coding, but they will probably do you little good until you've
chewed on some data yourself. I don't know of a "cookbook" methodology for
character coding anywhere in the literature. There are probably as many ways
of coding characters as there are systematists (and I have seen a paper to
that effect, believe it or not). Once you get familiar with it a little
more, you can go back and access some of the hard-core literature.

Hope that helped,