Fwd: RE: Re: [TAXACOM] More on the 'cladistics' of sequences

[pierre deleporte]

>A 16:02 10/06/2004 -0400, John Grehan wrote:
>>Pierre:
>>Thus they [character states] can be either convergent, or synapomorphies 
>>at a deeper node. Better enlarge your analysis.
>>John:
>>Yes they do require further consideration.

I don't mean that, I mean using the multiple outgroups analysis. Put more 
taxa in the analysis, including the diversity of putative outgroups for 
these ambiguous characters, and the ambiguity between the two options above 
will hopefully be resolved on the optimal cladistic topology.

>>J: My understanding is that the programs will recognize a level of 
>>congruence of this character with others showing the same relationship.

This is too vague... The program will tend to make similar character states 
close together according to cladistic criteria (maximizing homology by 
topologic contiguity), and root on your outgroup(s) criterion. The two 
operations are independent. Because of logics. Mathematical theory of 
graphs tells us that a cladistic topology remains a cladistic topology, 
whatever the way you root it. The same way a phenetic topology remains a 
phenetic topology, whatever the way you root it. Topology is a property of 
graphs independent from rooting. Logic matters. And it is first: no logic, 
no science, because one would be allowed to utter any inconsistencies 
without a safeguard.

>>J: Only it seems that some people do not polarize the characters a priori 
>>as they leave them unordered in the analysis.

This has logically nothing to do. Are you now confusing "ordered character" 
with "polarized character" ?
I can have the unordered multistate character as follows: states 0, 1, 2, 
3, and any change from one state to another one will cost the same.

Now I can have the ordered multistate character as follows : states 
0-1-2-3, and a change between two adjacent states in this ordered series 
will cost 1 point, a change between 0-2, or 1-3 will cost two points, and a 
change between 0 and 3 will cost 3 points (or any other weighting at your 
guise).

This is the meaning of the term "ordered" in cladistics. Ordering requires 
a special model of character evolution in which all intermediary character 
states must be passed through during evolution. Unordering requires another 
special model according to which all changes are equally plausible. But 
this has nothing to do with polarizing the character. The ordered character 
above is still not polarized in putative plesio-apomorphy.

Now you can putatively polarize these characters on state 0, or state 3, or 
other: the state present in the outgroup will be the putatively 
plesiomorphic state. Programs process the data just like that.

Now, if by "ordered", you mean "polarized" in fact, then just read again 
all the previous messages on the topic and please think a little bit about 
them:
those people using programs DO polarize characters, because they DO 
indicate the outgroup, which the program DOES use for polarizing the 
characters. Thus, all characters in the papers you read are a priori 
polarized (no obligatorily ordered) from the moment when an outgroup(s) is 
indicated.

I don't know how I can state it so that you pay attention. It's crucial 
concerning your misconceptions about molecular phylogenetics. They are 
clearly rooted in your misconception of what the programs do, and some 
aspects of what cladistic analysis is about.

To try and help Curtis Clark efforts:

When you say "phenetic character", John, other practicians of cladistic 
analysis will say: "character polymorphic in outgroups".
Not evident for mutual understanding...
Because when the other people say "phenetic", they mean "clustering taxa on 
the basis of overall similarity between their respective sets of character 
states" (e.g.: percent similar character states shared by two taxa or 
groups of taxa, and then two other ones, and so on...). Which of course is 
meaningless concerning one character alone (you can't compute a range of 
percent of shared similarities with only one possible similarity at stake; 
it can just be 1 or 0, period).

This is why nobody ever talks of a "phenetic character".

Now, "polymorphic in outgroups" is clear enough to state what you mean.
And molecular data are not particular relatively to morphological ones in 
this respect, because both kinds of data may happen to have, or have not, 
polymorphism in putative outgroups. Some morphological characters may 
appear to be evolutionarily versatile, and some molecular ones highly 
conservative.

Pierre


Pierre Deleporte
CNRS UMR 6552 - Station Biologique de Paimpont
F-35380 Paimpont   FRANCE
Téléphone : 02 99 61 81 66
Télécopie : 02 99 61 81 88