Talk:Parsimony and its role in Phylogenetic Reconstruction
From EvoWiki
Wikify
This article (and the talk page for that matter) needs to be wikified. It's hard to read a block of text. Are the original authors around to do this? I'm far too busy to read through it.--Doddy 09:40, 2 April 2007 (BST)
Other Comments
"As applied to phylogenetic analysis parsimony makes one explicit assumption: that homoplasy, parallelism, and character reversals, are all rare" -- this is just wrong syn. It assumes convergences/reversals are minimized vis-a-vis alternatives. A parsimonious tree can have an absolute ton of each. --GFA 19:46, 6 Feb 2004 (GMT)
I see it as a matter of almost semantic quibbling, but will change it. It alters the argument that naive adherence to parsimony is just that, not in the least. (JGK 11:44 GMT -5, 2/6/04)
Oh, but it does syn! It makes "there is alot of this stuff in vertebrate evolution" essentially a non-argument --GFA 04:53, 7 Feb 2004 (GMT)
The argument is still that the phylogeny that needs to invoke less convergence etc., is more likely to be correct, no matter how many times convergence crops up in the phylogeny being advanced. And thus, if we know that such patterns are rampant in evolution, we cannot dogmatically adhere to parsimony particularly in cases where the possible hypotheses are all very similar, with negligible differences. I am NOT arguing that parsimony is incorrect or useless, but only that we should be stringently applying it and not categorically rejecting a valid hypothesis for the simple reason that an alternative may be slightly more parsimonious. Nor do I accept sanctimonious preaching about "objectivity" as a valid criticism--there is no objectivity in systematics, and there never has been. The search for the magical formula by which systematics will be rendered as precise, sterile, and devoid of personal subjectivity as the hard sciences, is as elusive and farcical as the quest for the Sangreal. (JGK 12:20am GMT -5, 2/7/04)
Having reread this, syn, I fail to see how the discussion of Gaviomorphae supports the thrust of your argument; indeed, it would seem to argue just the opposite! Assuming for the moment that all but two of the characters are homoplastic, rejecting it for that reason would be, by golly, parsimonious!
The only sense in which it would count against parsimony is if there were some reason to accept is monophyly in spite of the homoplasy you say it requires; for example, if the true phylogeny were known to be otherwise.
If, on the other hand, its one of these "garbage in, garbage out" arguments, it has nothing to do with parsimony but rather with potential problems with character data. These are two entirely different things --GFA 10:29, 25 Feb 2004 (GMT)
It is a case in which parsimony dictated, multiple times, that "Gaviomorphae" or Gaviidae + Podicipedidae was holophyletic, for the simple reason that it was more parsimonious to believe their suite of specializations geared toward a diving ecology, had not evolved independently of each other. And though Mayr & Clarke (2003) concluded that character support for Gaviidae + Podicipedidae was weak, their strict consensus cladogram of ordered/unordered characters still recovered an obviously false phylogeny...via parsimony. And for some three decades repeated cladistic analyses invoking this same reasoning have defended the validity of one of these two clades. This is therefore an ideal example of a case in which parsimony dictates a clearly incorrect phylogenetic hypothesis and is thus exemplar of the inability of strict parsimony analysis to differentiate between inaccurate and accurate phylogenies in at least some cases.
Secondly, there is really no great ambiguity about it--these characters are homoplastic in these two groups. So there is no need to offer the assumption that this is the case.
What I really fail to see is how this has been construed as "anti-cladistics," as I have no major quarrel with cladistic methodology--I wrote the wiki's primer on it, e.g. What I do have a quarrel with is this idea that parsimony analysis is the infallible panacea for all systematic dilemmas, when so clearly it is not. And with it goes the inerrancy of cladistics. Whatever its systematic rigor, cladistics is as fallible and subjective as any other methodology.
JGK (7:30 GMT -5, 2/25/04)
Heres what im saying syn. To the extent that the characters are indeed homoplastic, and to the extent that its monophyly is rejected accordingly, you are being parsimonious. This is obviously because if monophyly was maintained in these other studies, it is presumably because a para/polyphyletic Gaviomorphae would require more homoplasy.
If you are to argue that Gaviomorphae is para/polyphyletic because it requires alot of homoplasy, thats fine, but doesnt help your central argument. If you want to argue that Gaviomorphae is para/polyphyletic and requires more homoplasy, that would work, but its unclear on what grounds you would argue it (why would you reject Gaviomorphae for the amount of homoplasy it requires, and yet uphold para/polyphyly, which would require more?). -GFA 20:03, 25 Feb 2004 (GMT)
I reject the holophyly of "Gaviomorphae" and indeed any clade linking Gaviidae and Podicipedidae, for a variety of reasons, not simply because either clade requires a high degree of homoplasy. "Gaviomorphae" can be rejected as holophyletic on the grounds that there is a fundmantal lack of homology in at least one major element of the hindlimb, the cnemial crest, which is composed of different elements in each of the constituent lineages of "Gaviomorphae" and not a single synapomorphy can be presented of this assemblage which cannot be explained in light of this lack of homology, as anything but convergent. The same rationale applies to the proposed Gaviidae + Podicipedidae clades. We know that they have at least one fundamental lack of homology in the structure of their diving apparatus, and we further know that there is practically no trait which can be advanced as synapomorphic of this putative assemblage which is not demonstrably convergent. Adding to the woes of the loon/grebe nexus is the molecular evidence that indicates: a) grebes have no close living relatives least of all loons (Sibley & Ahlquist), or b) grebes are near to Phoenicopteridae. So, we have a few things in this case:
a) A clearly non-existent clade in which all major potential synapomorhies are explicable as convergence
b) Fundamental lack of homology between at least one major component of their respective diving apparatus
c) Parsimony analysis consistently allying these groups as sister clades (in this case I mean Gaviidae + Podicipedidae) because it is the more parsimonious assumption.
Something is clearly very wrong here, since we are getting false conclusions from these cladistic analyses. Either the morphology is being tricksy and lying to us, or parsimony analysis is. Call me a heretic but my money is on our methods of analysis, not on reality, as the culprit.
JGK (7:40am GMT -5, 3/1/04)
Here are our options:
1) Gaviomorphae is para/polyphyletic because particular characters, e.g. the cnemial crest, fail to meet the similarity requirement of homology.
- This is not an argument against a consistant application of parsimony in particular, or against cladistic method in general. Assuming it is true, its a character data problem which can potentially infect any method of phylogenetic construction.
2) Gaviomorphae is para/polyphyletic because the characters used as support fail to meet the congruence requirement.
- Lets assume for the moment that the characters of the sort described above do in fact pass the similarity test and/or are excluded from the matrix. How then are you to say if they and the others are or are not homologous? On what grounds (again, leaving aside those that are not structurally similar) are you dismissing them as convergences?
First and foremost, there is no "assuming" that the cnemial crest in the constituent taxa of Gaviomorphae or Gaviidae + Podicipedidae are not homologous--they demonstrably are not. Second, I am not arguing against cladistic methodology in phylogenetic reconstruction, so naturally you will not find an attempt to do so in this article. Thirdly, it is perfectly relevant to the role of parsimony because parsimony anlaysis consistently groups these birds (nowadays this is true for a Gaviidae + Podicipedidae clade) as a holophyletic grouping in spite of the fact that is known to be a false phylogeny. The problem is not the morphology or the character matrix, it is the underlying assumption being used to analyze them.
In your second point, am I right in interpreting it as: "let us assume that the cnemial crest between these taxa is homologous?" If that is the argument you are offering it is an irrelevant example since we know that the cnemial crest between these forms is not homologous. If you are asking how it is that we know that this suite of characters associated with diving birds are convergent in general, I refer you to Storer. They are all present in a variety of diving birds which are not closely related, and thus are prima facie examples of convergence imposed upon a phenotype by shared stimuli.
JGK (9:24pm GMT -5, 3/2/04)
1) Parsimony, in cladistics, holds that the true phylogeny is likely to minimize homoplasy vis-a-vis alternative phylogenies.
2) If parsimony is an incorrect assumption in certain cases, it would be the case that the true phylogeny requires more homoplasy than the most parsimonious tree.
3) The minimization of homoplasy is a criterion independent of potential problems with character data, making the fact that "parsimony anlaysis consistently groups these birds" irrelevant. Parsimony is only one part of parsimony analysis (cladistics).
The question is, how does entering incorrect data into your matrix establish that true phylogenies sometimes require more homoplasy? It doesnt.
4) As per 2, arguments that clades require a large amount of homoplasy are obviously not arguments against parsimony.
5) On what grounds are you arguing that the other characters are convergent in this particular case? Does maxillary/dentary tooth loss in ornithomimosaurs and oviraptorosaurs establish that the state is convergent in neornithines as well? Of course not.
