Re: Transformational and taxic approaches to character coding [was Re: Philosophies for Character Ordering]

["Jonathan R. Wagner" <jonathan.r.wagner@mail.utexas.edu>]

As usual, do recall that I tend to get quite huffy when people reply to my
posts and don't cc me, because then I'm the last person to know about their
reply. :) Much as I don't fancy becoming further mired in this thread, I
can't resist:

Jaime A. Headden wrote:
>   As Mickey pointed out correctly, "ordering" argues that a taxon with
> condition "2" must have "1" as well as "0", [...]
> A taxon that makes a "jump" from "0" to "2" is inferred to have gone
> through stage "1" in its evolution (I made this point by bringing up where
> a taxon can incorporate two vertebrae into the sacrum simultaneously
> rather than one by one; the machine will interpret this the same way).

To clarify for the audience, what Jaime is talking about certainly COULD
happen (and probably has at least once in the history of life), and anyone
who says it cannot needs to look more closely into the development of the
sacrum. However, from a "taxic" perspective, the quesiton is not whether
this CANhappen, but whether or not you wish to include the (potentially
valuable) information encoded by the intermediate state. Do you feel that
"presence of six or more sacrals" should be a valid potential synapomorphy,
or do you want to sweep that potential under the rug? By running your
character unordered, you a priori assume that THE NUMBER OF SACRALS HAS NO
VALUE IN DETERMINING THE TREE EXCEPT IN CASES WHERE THE NUMBER IS EXACTLY
THE SAME. Remember, although we cound steps, the real determining factor is
not differences, but similarities.
    Here's where transformationalism eats its own tail: just because two
taxa have the same number of sacrals does not mean that they are the same
vertebrae (not an issue for taxic thinking). Developmentally, growth of the
ilia induces the "sacralization" of vertebrae: if two taxa form an ilium of
about the same length and structure, they can sacralize the same number of
vertebrae... but one could do it quite a few segments up from another,
sometimes in the same embryo (hence the oft-cited bilateral assymetries
seen, for example, in lizards and frogs). From a transformational
perspective, can you really risk coding these as the same character? Or take
_Hypsilophodon foxii_, with its "sliding" sacrum... the sacrum incorporates
slightly different vertebrae, and different numbers of vertebrae, in
different individuals. How would you code this, as a transformationalist?


Tim Williams wrote:
> What about (say) the transformation of caudals in titanosaurs: (1)
> procoelous; (2) amphicoelous; (3) opisthocoelous.  A given caudal may have
> switched from (1) --> (3) without passing through (2).

    Yes, it may well have, historically. Actually, from a taxic perspective,
this is not the best case for ordering. The states do overlap, in that each
condylar morphology shares a cotylar surface with the intermediate state,
but there is also overlap among the end-member states, specifically,
presence of a cotylar articular surface. I would ordinarily code something
like this as a series of characters, shown below. Note that I specifically
code for the presence of cotylar caudals as a similarity among these
organisms, even though the picky anatomical details (i.e., which end is
cotylar) are different.

1. Caudal vertebrae: 0) amphicoelous, 1) with condylar articulations on some
vertebrae
        If the condylar articulations extended throughout the column in some
taxa, I would add a third state for that, and order it because the states
CLEARLY overlap.
2. Caudal vertebrae with condylar articulations: 0) on proximal face of
centrum, 1) on distal face of centrum.
        This character must be scored as unknown for taxa that have
character 1, state 0.
    Note two things: this potentially leads to the old feathers
red/blue/absent problem, but, as pointed out by others, that is probably
better than not coding the information at all. Also note that I am NOT
subscribing to the convention that the ancestral state must be coded as 0.
Things get a more complicated if you have biconvex vertebrae (as in crocs)!

An alternate coding:
1. Caudal vertebrae: 0) amphicoelous, 1) with condylar articulations
2. Distal articulation: 0) cotylar, 1) condylar
3. Distal articulation: 0) cotylar, 1) condylar
...is less  promising, because the states of each character seem logically
non-independent (tranformationally, if one changes, at least one of the
other two *must* also change). The transition costs (an application of
transformationalist thinking) are also telling: going between morphologies
has a cost of two, giving the arrangement of caudal vertebrae two "votes" in
the final tree. This might be a reasonable choice under some circumstances.
However, since there is logical non-independence, I would try to find a
simpler coding (as in the one shown above). I might consider some of these
characters in the presence of biconvex vertebrae.

Another alternate coding is arguably even more dodgy:
1. Caudal vertebrae: 0) without condylar articulations, 1) with condylar
articulations
2. Caudal vertebrae: 0) procoelous, 1) amphicoelous, 2) opisthocoelous
[UNORDERED]
    Again, note the 1:0 is logically non-independent of 2:1, and thus
amphicoely gets two "votes" in determining the best tree, but the other two
morphologies get one.

However, again from a taxic perspective, if one coded the character as Tim
mentioned:

1. Caudal vertebrae: 0) procoelous, 1) amphicoelous, 2) opisthocoelous
[ORDERED]

...so what? The analysis will require two steps to make the transition from
pro- to opisth- (again, in my opinion this ignores the value of the
observation that condylar articulations are present). From a taxic
perspective, this is not necessarily the same thing as saying that the
lineage actually DID pass through that state, only that the ability of the
intermediate state to group the taxa differently has been overridden by the
data. Hopefully, you would *have* enough data that phylogenetic signal will
override this problem. You can also selectively unorder this character to
see if the relevant parts of the analysis are sensitive to that
"assumption." In my limited experience, such sensitivity is rare.

My two cents.

Wagner