[Taxacom] Zander's point on linked-traits

Kenneth Kinman kennethkinman at webtv.net
Wed Feb 3 12:18:53 CST 2010

Dear All,
     First let me remind John Grehan that I do NOT oppose cladistics,
and never have (it is mainly strict cladism's attack on paraphyletic
taxa that I oppose).  My goal is to make cladistics better and more
reliable, and Richard Zander brought up an important point.  Parsimony
can fail to find the right answer if a bunch of linked traits
(conceivably even the result of a single mutation in a regulatory gene)
are each weighted equally with more informative unlinked traits (the
latter thus being numerically overwhelmed, resulting in a severe
dilution of valuable information).         
     In a couple of Taxacom posts back in April 2003, I warned John
Grehan that many of his orangutan-human "synapomorphies" seemed
clustered in body regions with regulatory links.   I still believe that
this will eventually explain why he is finding such
"pseudo-synapomorphies", which would thus actually be homoplastic
(plesiomorphic or convergent).  And ANY morphological cladistic analysis
which contradicts the findings of many different molecular cladistic
analyses should be evaluated for this problem.  You can read my concerns
in the following Taxacom posting:


Richard Zander wrote:
And . . .  how many cladistic analyses are checked to see that atomized
traits, no matter how homologous, may not be probabilistically analyzed
in many cases. Cladistics assumes that:  when you have 3 traits
supporting (AB)C and only 2 supporting (AC)B, then chances are the
shared ancestor had the three traits. But suppose the three traits are
selected for as a package, i.e. all three are necessary to survive in a
particular environment, and the other two are unlinked. Then, chances
are (AC)B is correct. Morphological analyses are never checked as to the
degree of linkage of traits in each species in the data set in any
adaptive environment. 

More information about the Taxacom mailing list