[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: Polytomies in cladistics: Good, Bad, or Ugly?



----- Original Message -----
From: "Phil Bigelow" <bigelowp@juno.com>
Sent: Thursday, September 15, 2005 4:50 PM
Subject: Polytomies in cladistics: Good, Bad, or Ugly?

In cladistics, are polytomies (3 or more branches from one node)
considered to be a valid working hypothesis, or are they instead viewed
as an artefact of a limited data set, weaknesses in the algorithm, or a
bad choice of test parameters?

They are never seen as the result of weaknesses in the algorithm.

This leaves three possible interpretations of polytomies:

(1) The true simultaneous split of a population into several. (In principle this is easy to imagine -- for example, take the classic model of allopatric speciation and suppose that two or more barriers arise instead of one.)
(2) A very fast succession of two (or more) dichotomies -- so fast that no detectable apomorphies had time to evolve between the first and the second (and so on).
(3) A mere lack of data. (Like (2), but replace "detectable" with "detected".)


The third is by far the most common assumption. After all, it is very easy to make a too small data matrix!

One famous polytomy was widely thought to be a case of (1) or (2): the split between humans, chimpanzees and gorillas. For 20 years or something most molecular studies found a trichotomy, while others found all three possible resolutions.* Then people turned to longer sequences of faster-evolving genes, and it turned out to be a case of (3).

What do you mean by "bad choice of test parameters"? Things like maximum parsimony vs Bayes*? Indeed some methods are less sensitive to artifacts like long-branch attraction than others, and some or all of those artifacts can produce polytomies. But this is actually a case of (3) -- more robust methods simply need fewer data to get a resolved tree.

* I'm not sure if any molecular study found chimps and gorillas as sistergroups, but morphology did anyway.
** There is a paper on Bayesian inference for morphology (applied to trilobites) in the latest Systematic Paleontology. I haven't read it yet.