[Taxacom] very nice opinion article in today's Zootaxa

Richard Zander Richard.Zander at mobot.org
Fri Sep 30 16:04:01 CDT 2011


I'm of course not discussing, Michael, the tail end of species with only
a few known specimens, which are described on analogy with similar
better known species.

In my field, there are many more than 1000 specimens collected world
wide for all the broadly distributed species. These have been studied by
many taxonomists over the past 250 years (building on each other's work)
yielding pretty good understandings of the distribution of traits, most
unimodal, some with subspecies being multimodal, etc. 

Okay, for an extreme test, let's say some super-annuated near-sighted
wizened taxonomist is doing a revision and analyzes (measures and
otherwise squints at traits to contribute to a description that includes
trait variation) only 100 specimens of each species. Forget that the
specimens she is looking at have already been examined by others. 50
species X 50 traits X 100 specimens = 250,000 data points, which is five
times size of the data set of the molecular work (50 X 1000). Divide by
half to get just conservative traits, and you get 150,000 data points,
pretty good for a taxonomic method accused of having few data compared
to molecular systematics.

Formalizing the heuristics of alpha taxonomy is an important thing to do
in the future, and 100 samples may not be enough to accurately gauge
distributions (see Smith, Z.R. & C. S. Wells. 2006. Central Limit
Theorem and Sample Size. Paper presented at the annual meeting of the
Northeastern Educational Research Association, Kerhonkson, New York,
October 18-20, 2006. You can find it on the Web.) 

Re your response, alpha taxonomists (doubtless including both of us)
commonly list only representative specimens examined. Most do not have a
statistical understanding about how and why their methods work, I mean
as opposed to "statistical phylogenetics" which does not deal with
distributions of specimens in almost all cases, an when they do, 49 is
considered a huge sample for a single species. And yes you have seen
that done, Michael, taxonomists do build on the work of others. And yes
you have done that, every time you describe a new species by analogy
evaluating similar traits in similar taxa against the background of
known species with well-known trait distributions (variation) based on
thousands of observations worldwide (well, ideally).

R.

* * * * * * * * * * * *
Richard H. Zander
Missouri Botanical Garden, PO Box 299, St. Louis, MO 63166-0299 USA  
Web sites: http://www.mobot.org/plantscience/resbot/ and
http://www.mobot.org/plantscience/bfna/bfnamenu.htm
Modern Evolutionary Systematics Web site:
http://www.mobot.org/plantscience/resbot/21EvSy.htm


-----Original Message-----
From: Michael A. Ivie [mailto:mivie at montana.edu] 
Sent: Friday, September 30, 2011 3:39 PM
To: Richard Zander
Cc: James Whitfield; taxacom at mailman.nhm.ku.edu
Subject: Re: [Taxacom] very nice opinion article in today's Zootaxa

Hmm, Rchard, looking through your publications, I do not find one on
multiple species with a minimum of one thousand specimens per species
listed in the material studied.  Did I miss it?  I certainly have never
accomplished that level of specimens in my material studied!  In fact, I
don't think I have ever seen that done.

Michael Ivie
>
> Compare:
>
> A molecular data set of say 50 species, represented each by one 
> exemplar, and 1,000 phylogenetically significant sites gives 50,000 
> data points. This is considered a big data set.
>
> But, a morphological data set of 50 species may have each 50 traits 
> represented (over 250 years of study determining trait distributions) 
> by an average of 1000 specimens each, and we get 50 x 50 x 1000 = 
> 2,500,000 data points. Say only half the traits were genuinely 
> conservative (resistant to adaptive change) and we get 1,250,000 data
points.
>
> The difference is that the molecular data set results in a cladogram 
> that does not cluster specimens. The 95% credibility distinguishes 
> single specimens, so one can easily imagine two or more samples 
> occurring in one broad morphological distribution of traits. Just 
> because the two samples differ at 95% or more in credibility is no 
> cause to postulate two cryptic species.
>
> Only if the distribution of molecular variation of many samples 
> matches the distribution of morphological variation of many samples is

> there genuine support of molecular data and morphological data. If 
> there are only small samples of molecular data, any match is mere 
> corroboration ("not incompatible with") not support, for the 
> morphological study. A mismatch means nothing.
>
> The burgeoning number of samples per taxon in barcoding is excellent 
> news.
>
> * * * * * * * * * * * *
> Richard H. Zander
> Missouri Botanical Garden, PO Box 299, St. Louis, MO 63166-0299 USA 
> Web sites: http://www.mobot.org/plantscience/resbot/ and 
> http://www.mobot.org/plantscience/bfna/bfnamenu.htm
> Modern Evolutionary Systematics Web site:
> http://www.mobot.org/plantscience/resbot/21EvSy.htm
>
>
> -----Original Message-----
> From: James Whitfield [mailto:jwhitfie at life.illinois.edu]
> Sent: Friday, September 16, 2011 5:36 PM
> To: Richard Zander
> Cc: Doug Yanega; taxacom at mailman.nhm.ku.edu
> Subject: Re: [Taxacom] very nice opinion article in today's Zootaxa 
> Not to argue for using only one sample to stand for a species, or only

> one DNA barcode, as I agree with you that it is a bad idea.
>
> But if people would bother to look carefully at the original papers, 
> they are based on LOTS of molecular samples from each species.  It is 
> the CONSISTENCY of the molecular or morphological differences that is 
> critical, not how major the differences are. In this case the 
> molecular and morphological differences are consistent across many 
> samples within putative species and do indeed support one another.
>
> Most species descriptions based on morphology (probably 95% + of all 
> species descriptions?) do not have any convincing analysis associated 
> with them. We are in the habit of trusting the experts in this case 
> rather than requiring strong evidence.  Most of the time this works 
> really well, I'd say.
>
> Jim




More information about the Taxacom mailing list