keys are horrible

Mr Fortuner connection modem fortuner at MATH.U-BORDEAUX.FR
Sat Sep 9 07:33:03 CDT 1995

While some of the key examples are funny (I love the one about the Dutch bird
- but what if the bird makes an identification error and sings the name of
another bird?), others demonstrate, once again, that there is more to
identification than keys. Going back through a key or using probabilistic
("usually") statements are things that would be far easier to do with other
methods: when you calculate a coefficient of similarity, you don't need "not
as above" statements and if you want to do backward chaining, use an expert
system. Keys are great for eliminating species, but other identification
functions, e.g., resemblance and recognition, are just not possible with

The other problem I have seen with keys is that their authors tend to
follow the classification. You have to use a family key to find the right
genus, then a genus key to find the species. This may not work. For example, a
nematode subfamily was created some time back by splitting a large genus
(Criconemoides) into a dozen genera. This taxonomic decision was justified,
but the resulting genera are very difficult to separate from each other
because the phylogenetic characters used for defining them are very difficult
to observe. Using a key to the genera often results in identification failure.
Many identificators use an old key to Criconemoides species, find the species,
then find to what genus it now belongs.

There is no reason why the
identificator should retrace the steps of the classificator, and we should
keep separate in our mind identification characters and taxonomic/phylogenetic
characters. Some "synapomorphic or autapomorphic features" may be difficult to
observe and using such characters may result in identification errors.

I have
proposed a few years back the concept of "nest of species" based on "primary
identification characters", which are characters that are "easy" to record. A
nest includes all the species that have the same set of primary identification
characters. In my example, the old genus Criconemoides would become the nest
criconemoides, with low-case "c" and no italics. The point is that, by
definition, nest identification is based on "easy" characters, which means
that any moron can do it. Species identification would still have to use
secondary (not easy) identification characters, but at least, the
identificator would be reasonably sure that he is working within the right
group of species. Some species could be eliminated (using the remaining
primary characters), then the identificator would have to switch to other
methods to find the most probable answer.

Renaud Fortuner

More information about the Taxacom mailing list