Darwin (was: Phylogenetic evidence)

[Ken Kinman]

Pierre,
       Trying to resolve a trichotomy (like BCD) with just three characters
would be rather scary (the more characters there are, the better I feel).
But if I did have just three characters to go on, I might try resolving the
trichotomy by weighting the characters if possible (although I know this
makes some cladists extremely uncomfortable).
      For example, if I were looking at three primitive bird families (all
based on relatively complete skeletal material), I would give more weight to
the presence of a pygostyle than I would to the absence of teeth.  Both of
these are derived characters for early birds, but fusing vertebrae into a
pygostyle probably occurred rarely compared to the loss of teeth.  If the
third character was also relatively strong and congruent with the presence
of a pygostyle, I would go with those two characters and just ignore the
absence of teeth.
     With so few characters to work with, I think weighting of some kind
would be particularly important.  Thank goodness we usually have more than
just three characters to work with.  It would be especially bad if you had
one moderately strong character and the other two were weak (and very
subject to homoplasy).  In that case, I would probably just stay with an
unresolved trichotomy---or better yet, look for other characters.
          --------- Cheers,  Ken
*****************************************
Pierre Deleporte wrote:
>A way to deal with this problem is to consider the "three taxon analysis"
>(3TA) debate in the recent literature.
>
>Nelson (1996, "Nullius in verba") wrote :
>
>"Several years ago, J.S. Farris and I agreed to disagree over the
>interpretation of a data matrix for taxa ABCD, with a trio of hypothetical
>characters conflicting in all possible combinations of 2 of 3 taxa BCD:
>
>Matrix 1:
>
>O  000
>A  000
>B  110
>C  101
>D  011
>
>With inclusion of an all-zero outgroup (O) in the data matrix, Hennig 86
>yield six trees : three with a basal trichotomy, AB(CD), AC(BD), AD(BC),
>and three resolved tres, A(B(CD)), A(C(BD)), A(D(BC)), all with length 5,
>ci 60, ri 33. The strict consensus of the trees is uninformative.To me,
>however, it seemed that the trio of characters might be seen to associate
>taxa ABCD, such that only the resolved trees, and their informative strict
>consensus A(BCD), are relevant. Farris stated that such should simply not
>be don. The reason was that his optimization criteria produce no
>unequivocal synapomorphy for the group BCD."
>
>The following debates are quite interesting regarding the point at stake.
>It appears that Nelson resolves the phylogeny A(BCD) using his "three taxon
>analysis" (3TA) procedure.
>It appears that this procedure is indifferent to topographic contiguity =
>homology by descent, and also to reversals as possible synapomorphies.
>It appears that a modified (improved ?) version of it (m3TA) leads to plain
>classic phenetic clustering (by overall similarity based on all occurences
>of traits considered separately two by two, absence excluded).
>
>To the objection: "why should all these characters partly point to group
>(BCD) if there is no such a group ?", it was answered: "why can't we find
>the slightest unambiguous synapomorphy if such a group really exists ?"
>
>The conclusion is that standard cladistic phylogenetic analysis, quite
>logically, finds no resolution without some non ambiguous synapomorphies,
>and that competing so-called "cladistic approaches" find such "cluster
>resolutions" only when... they inadvertently "rediscover" and apply
>phenetic methods.
>
>The charge of clustering on similarity instead of defining classes can be
>addressed to 3TA, and to m3TA = phenetics, not to standard cladistic
>parsimony analysis. Reversals is not a problem, and the "classic" examples
>in this debate merely do not contain cladistic resolution.
>
>For nomenclatural purposes, opting for naming a group (BCD) in this case
>would be applying "locally" a phenetic clustering, forcing taxonomic
>resolution, possibly in an otherwise classic phylogenetic system. No
>problem for me, a cladist without a resolved cladogram should not object
>for alternative expedients... provided that warnings should be clearly
>posted: "This is a local phenetic clustering, not to be taken as resolved
>phylogeny". Maybe a "reasoned eclecticism" is possible this way, with clear
>priority rules for methods and criteria... (should please to Ken, some
>way...).
>
>If this can help...
>
>(The whole "3TA" debate can be followed mainly in Cladistics and Systematic
>Biology of the ten last years, notably papers by Nelson, Platnick,
>Scotland, Carine, Williams, Farris, Kluge, de Laet, Smet, Deleporte, and
>others... sorry for those I don't remember now...).
>
>Pierre




_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com