(Further thoughts on problems I see in structural decompositions).

The main unknown problem I see is how to express variability and lack or uncertainty of knowledge about structural observations. This may be exceptional in the extremely well studied higher plants, but not so in small organisms that are often only imperfectly studied yet.

An example is "Chlorophyll in stems or leaves". You basically suggest reformulating it to fit a given model that is considered more effective for identification. In higher plants the assumption would be that since the information is not available in the combined statement, this involves some literature research, and perhaps looking at a herbarium specimen. However, already in higher plants in a key to families this is far from trivial, and in many organism groups this imprecise synthetic statement may be the only knowledge available.

Very similar case occur very frequently for temporal or spatial modifications (which may be "regions" or not).

Also, I think you suggest in addition to Leaf-Cholorphyll Presence and Stem-Chlorophyll Presence to have:

Entire Plant - Colour - specify any colour

Which I believe would truly have to be above-ground vegetative parts (I assume that the root color would not be considered, although roots are included in entire plant). Providing these combined sub- partitions (do you do that in the Prometheus model?) is necessarily a choice among multiple combinations possible (for vascular bundle characters combining root and stem may be desirable) that can not be fit into a single hierarchy (do you provide multiple hierarchies?).

I do agree that better structured models have analytical advantages. So far our believe is that currently no better model is well tested and that limitations in the range of possible statements (which naturally do exist in the SDD model as well) need careful consideration and testing. In that case the attempt of Prometheus to do exactly such testing (i.e. not only to design an ontology model, but record data and test its use in identification processes is extremely valuable for the future.

My own goal would be to allow less structured descriptions being expressed in SDD, but to support working ("voluntarily") in a model that is highly and systematically structured. Applications may have "modes" for rigourous or lenient behavior, but not the exchange format.

Gregor ---------------------------------------------------------- Gregor Hagedorn (G.Hagedorn@bba.de) Institute for Plant Virology, Microbiology, and Biosafety Federal Research Center for Agriculture and Forestry (BBA) Koenigin-Luise-Str. 19 Tel: +49-30-8304-2220 14195 Berlin, Germany Fax: +49-30-8304-2203

Often wrong but never in doubt!