[Taxacom] Read... and believe...

[Mike Dallwitz]

Stephen Thorpe wrote:

"I certainly think you are correct inasmuch as lack of comparative 
descriptions is a problem in taxonomy ... but I don't think it is strictly 
relevant to the subjectivity/objectivity issue under discussion here."

A good description will help others to understand and reproduce an 
author's concept of a taxon - i.e. to identify specimens as belonging or 
not belonging to the taxon. This is the process that I am suggesting 
should be made more reproducible/reliable/objective. This should provide 
the infrastructure for the possibly subjective questions of whether new 
taxa should be defined (or existing ones merged).

Bob Mesibov wrote:

"I'm having a little trouble understanding how you project your wish for 
well-organised comparative data over time, i.e. over the many decades that 
a particular group has been (and will be) studied. So, the comprehensive 
organising of character data works fine at Time N, but at Time N+1:

a) Old methods give consistent and reliable character states, but with new 
methods, you find that many of the states were incorrectly recorded.

b) Individual characters are shown to be more complex than previously 
thought, and require reworking as multiple characters with new and 
different states, and the Time N set needs to be chucked out.

c) Many new characters are discovered which have to be added to descriptions.

d) Some old characters really need to be discarded, e.g. they were 
believed to be apomorphic but are now known to be plesiomorphic.

The paper you quote IMO confuses inadequate description with outmoded 
description. I think those are two different things. One specialist I know 
bravely created a matrix with many more characters than any one previous 
description had ever included, then went back to types to fill in the 
states for a *lot* of species. (This was for a phylogenetic analysis.) He 
dealt with all 4 of the time-related issues above, but he isn't naive 
enough to think that his matrix will be 100% correct, or even useful, in 
100 years' time."

I don't see any reference to the time factor in the quote. The 
descriptions won't improve with time, but the point is that they are 
inadequate on the day they are published.

Your points (a)-(d) refer to characters. But the worst part of the problem 
is that most descriptions don't use characters (sensu Colless, D.H. 1985. 
On “character” and related terms. Syst. Zool. 34: 229–233). They just use 
words strung together. I'm not implying that the descriptions have to be 
in the form of a coded matrix. But the characters must be explicitly 
defined, and must obey the rules, e.g. mutually exclusive states.

(a) and (b) imply that characters can change. A properly defined character 
can't change. If, for example, a state is added, or the meaning of a word 
is changed, that produces a different character. A data matrix based on 
properly defined characters will be just as correct, and just as useful, 
in 100 years time as it is today.

(c) Yes. Or at least they _may_ need to be added. Depending on the 
circumstances, the old ones may be adequate indefinitely.

(d) No. A character can't be apomorphic or plesiomorphic; those terms 
apply to character states. If you discarded all the characters that had 
plesiomorphic states, you would have few if any characters left. You may 
_choose_ to discard old characters, e.g. if ones that are better (in some 
sense) are found, but the old ones retain their usefulness for 
identification, and may conceivably be better for that purpose than the 
new ones.

"This was for a phylogenetic analysis." Then it doesn't really matter 
whether the data are reproducible; the odds are no one will ever know or 
care. _Identification_ (or its equivalent) is the true test of 
reproducibility. Reading a description and comparing it with a specimen is 
_not_ equivalent to identification; it's too easy to agree with the 
description (like a leading question).

People who produce general-purpose data matrices or keys are at least 
trying to produce comparative and reproducible descriptions. But they 
generally don't succeed very well - a failed identification means that the 
key user couldn't reproduce the data (provided the specimen does belong to 
one of the taxa in the matrix or key).

The method I proposed in an earlier posting 
(http://mailman.nhm.ku.edu/pipermail/taxacom/2009-August/030098.html) 
tries to nip reproducibility problems in the bud.

"I'd like to stress the importance of illustrations (drawings, photos, 
SEMs) in taxonomic publications."

No argument about that. Most modern interactive-key programs allow for the 
use of images - but some do it better than others. For example, Intkey 
(typically) illustrates a character with a single image containing 
illustrations and text of all the states, displays this image 
automatically when the character is selected, allows state selection 
directly from the image window, and closes the window automatically after 
state selection. There are flexible options for the display of taxon 
images. For example, if a few taxa remain in an identification, the user 
can display, automatically tiled, all illustrations of a specified kind or 
kinds for these taxa, e.g. the genitalia and/or wing images.

"the humans most likely to read my papers couldn't care less whether the 
text descriptions were machine-readable and machine-analysable."

But if they use the descriptions (as opposed to the illustrations) at all, 
they presumably like them to be human-understandable (i.e. reproducible).

"No, it isn't taboo, it's just more complicated than you seem to think."

Well, it seems to be hard to get a response on Taxacom to postings on the 
subject. Are you suggesting that Taxacom subscribers don't like discussing 
complicated subjects?

-- 
Mike Dallwitz
Contact information: http://delta-intkey.com/contact/dallwitz.htm
DELTA home page: http://delta-intkey.com