How is classification started?

Here's my problem. Classification, these days, in systematics is not founded upon morphological or functional similarity so much as on the residues of history. In philosophese, these groups are not made by equivalence classes or similarity relations, but by processes that we recover, as best we can, from the characters or traits of organisms living or dead.

But these characters or traits (there is a difference - characters are abstact representations of biological facts about organisms; traits are actual physical features of organisms - ah... but which ones?) have to be chosen. They get put into an "instance matrix" for processing by the various black boxes of systematics, but if you add homoplasious characters, or characters that have been arrived at independently by different pathways in evolution, you will get a false "signal" of the past. All this is discussed in the standard texts; I suggest Elliot Sober's Reconstructing the Past (MIT Press 1988) as a primer to the philosophical issues.

So, here's my problem: how are the characters, particularly at the molecular level as molecular-based phylogenies become the predominant mode of classification, arrived at? I once asked an ichthyologist of note how he chose the characters for this purpose (we were talking about morphological characters, but the point remains), and his reply was, "Well, I should damned well hope I knew my organisms by now!" (or words to that effect). And he should, and so should all systematists, but if we want to avoid the "Cynical Species" problem ("A species is whatever a competent systematist calls a species", see McOuat's paper for why this is not so cynical as it has been painted), we need to know how these characters/traits are selected.

In short, how are they classified, so that they can be compared to recover history? I am not asking how homoplasy is identified compared to homology - that's a fish of a different river - but how classification proceeds before systematics, that is, classification of taxa, begins.

I thought to find some comfort and information among the bioinformatics literature. But all I found was reference to classifying by function or structure. Most structural classification of molecules appears to be done on the primary sequence, that lineal series of bases or monomers. Functional classification relies, by contrast, on the secondary, tertiary and later structures, as the molecules fold, because the functional role of a molecule in a cell or organism depends on the attachment sites it has for other biomolecules, and that depends on the shapes the three dimensional (and, if it changes over time, four dimensional) molecules form.

But classifying molecules in terms of their structure gets us back to the older problem of classifying by morphology, only now it is the morphology of molecules, and the consequent problem of molecules being that way for reasons of simple form - it can't fold any other way no matter what its past history is. And classifying molecules in terms of their function is context-dependent - in one organism a molecule has one function because other molecules have the shapes and parts to play in the cycles of biology in that species; in another species, it may have a different function. It's a bit like teeth. History can be lost or confused if we classify by function.

More on this later. Joe Felsenstein just gave me some refs to chew over...

McOuat, Gordon. 2001. Cataloguing power: delineating 'competent naturalists' and the meaning of species in the British Museum. The British Journal for the History of Science 34:1-28.

Sober, Elliott. 1988. Reconstructing the past: parsimony, evolution, and inference. Cambridge, Mass.: MIT Press.