Identification methodology

[Mike Dallwitz]

                                                                 21 March 1995

> From: Renaud Fortuner <fortuner at math.u-bordeaux.fr>
>
> Rigor in Delta? I believe a better term would be rigidity.

I don't believe you could substantiate this. The DELTA programs are very
flexible in that they have large numbers of options which may be combined in
various ways to serve a wide variety of purposes (sometimes even ones that I
had not envisaged). In fact, this flexibility is one of the major sources of
criticism of the programs, particularly of INTKEY. We have addressed this
problem by providing the `simplified' mode of operation in the Windows version
of INTKEY.

If you just mean that the programs don't have all the features that you and
other users might want, then I agree with you. We are working on it.

> In any cases, I would like to ask a couple of question to Mike Dallwitz.
> When we met in 1990 (at the ARTISYST workshop in California), I asked you
> how I could enter intraspecific variability using Delta (ISV, the percentage
> of individuals in a species that exhibit a particular state of a character)
> and you said that this type of information could be added to a "comment"
> field. Then I asked about metadata and you said to put it in the comment
> field also, which would make it kind of crowded.
>
> The question is, is this still true or have you modified the Delta standard
> to provide for ISV and metadata?

The current version of CONFOR (our main program for processing DELTA data),
allows the use of `inner comments' for metadata, e.g.
    4,2<at the tip <material from WA not yet seen>>
These can be omitted from natural-language descriptions at the option of the
user. The problem with storing probabilities in comments, e.g.
    4,2/1<4%>
(i.e. for character 4, 96% of individuals have state 2 and 4% state 1) is not
so much the `crowding' of information (which is no worse than in conventional
printed descriptions), as the fact that the current software can't use the
information except in natural-language descriptions. The planned new CONFOR
will interpret this information, and various other types too, and will be able
to hide or display the various types of information at the request of the
user.

Information about the new DELTA data format can be found in the files
newdelta.txt and newdelta.ps at
    ftp://muse.bio.cornell.edu/pub/delta/standard
The new types of information are still incorporated in the syntax of
`comments', for compatibility with the present format, and to make it simpler
for programs to ignore the extra information. This need not concern the user,
who will not normally look directly at the DELTA-format files.

> I see in your list of requirements that a program should give "Similarities
> between taxa. Whether the program can find the similarities between members
> of a set of taxa."
>
> Excellent, but does this mean just the list of characters that are
> identical in the unknown and a particular species, or are we talking of
> computing a coefficient of overall similarity?

INTKEY has some flexibility in that the user can specify whether, for example,
`state 1 or state 2' matches `state 2 or state 3', but does not compute the
similarity measures normally used for phenetic analysis. Another of our
programs, DIST, calculates a distance matrix, and yet another, NSIM, displays
this information in a readable form, listing the nearest neighbours of each
taxon.

> More generally speaking, is Delta still doing dichotomous keys only (or
> variants such as multiple entry keys) or have you moved to other methods of
> identification?

By `dichotomous' I presume you mean `deterministic' or `non-probabilistic'. In
that case, the answer is `no'. I don't doubt that probabilistic methods are
important or essential in some fields. I would like to incorporate them in
INTKEY, and have thought about ways of doing so. However, we must set
priorities, and I think that most taxonomists prefer to use deterministic
methods if possible. (For one thing, the data are easier to gather.) It's
generally more important to cope with the possibility of errors in using
deterministic data, which is done by INTKEY's error-tolerance mechanism. For
some further thoughts on this, see

Dallwitz, M. J. (1992). A comparison of matrix-based taxonomic identification
systems with rule-based systems. In `Proceedings of IFAC Workshop on Expert
Systems in Agriculture', pp. 215-8. (Ed. F. L. Xiong.) (International Academic
Publishers: Beijing.)

> Even more generally, I am curious to know what people think of dichotomy. I
> am asking Taxacom recipients to answer the little survey below directly to
> my address. I'll compile the results and post on Taxacom in about a week.

I look forward to seeing the results of the survey, but I think there is a
danger of a `silent majority' effect.

Mike Dallwitz                                  Internet md at ento.csiro.au
CSIRO Division of Entomology                   Fax +61 6 246 4000
GPO Box 1700, Canberra ACT 2601, Australia     Phone +61 6 246 4075