[Taxacom] New lizard species

[Kim van der Linde]

I just read the article, and I have mixed feeling about the article.

1. The central idea is that we can determine species limits by 
estimating where solid differences in the DNA can be found, as estimated 
using specific Bayesian methods.

2. The species concept used here is a strict phylogenetic species 
concept. Whether or not this corresponds with species using the 
biological species concept is unclear.

3. The number of individuals used for some populations is as low as one. 
if you have a single outlier. Humm, so much for an estimate of 
population variance, which is at the heart of the idea.

4. The use of guid trees is problematic at least. Why 4 species, and not 
5, 6 or 3. If you a priori start with 4, you of course can find four. As 
long as you have well-supported nodes with a posterior probability of 1, 
you can determine species. The fact that the 10-species guide tree 
provides FIVE species should have resulted in major alarm bells.

5. They use the SAME data to first estimate the species, and then to 
estimate the probability of those species. This is a circular process. 
You need two separate data sets to do this. This sounds way to much to 
self-conformation.

Overall, I am not buying this article. Maybe when it is done properly, 
with sufficient number of individuals per population, independent data 
sets and a search algorithm that determines what number of species is 
actually explaining the data the best, I might be persuaded to accept 
these phylogenetic species, in the context of phylocode.

Kim


On 6/4/2010 8:19 PM, Bob Mesibov wrote:
> If you haven't seen it, do read
>
> Leache, A.D. and Fujita, M.K. 2010. Bayesian species delimitation in
> West African forest geckos (Hemidactylus fasciatus). Proceedings of
> the Royal Society B. Published online as doi:10.1098/rspb.2010.0662
>
> In a well-argued and very clear paper, the authors plump for
> probabilistic determination of species, and name four *Linnean*
> species on probabilistic grounds. They say '...we are not aware of
> any morphological or ecological characteristics that differentiate
> these lineages,' and therefore diagnose their species as follows (one
> example):
>
> 'Diagnosis. This species includes all populations that cluster with
> those from the southern portion of the Congolian rainforest included
> in this study (southern Cameroon, Gabon and Congo), with strong
> support in the Bayesian species delimitation model.'
>
> I don't actually have a problem with the methods used or the
> argument. My concern is that the authors confuse Linnean species with
> lineages. Lineages can be diagnosed within a probabilistic framework.
> They give us an insight into how evolution might possibly have
> happened. Linnean species are those nice, easily recognised things
> that enable us to retrieve biodiversity information, write field
> guides and feel that we have a handle on natural history.
>
> It's been obvious for a long time that Linnean taxonomy and molecular
> taxonomy were headed in different directions, but this paper (IMO) is
> a clumsy attempt to yoke them back together.

-- 
http://www.kimvdlinde.com