Archive for May, 2019

What good are taxonomic ranks?

It is sometimes said that taxonomic ranks – genus, family, class and so on – are useless and should be discontinued.

(See for example: Animal Diversity Web, Phylocode)

I agree that in isolation, saying that a taxon “is a class” or “is an order” tells you nothing. One often sees statements like “90% of all families went extinct” as if “family” actually meant something. It doesn’t, because whether a group is designated a “family” or “class” is an almost arbitrary choice of the individual preparing a classification. The choice of rank may persist because taxonomists respect precedent, but when it’s assigned in the first place or under revision, it’s a free for all.

Sometimes people try to associate rank with historical divergence time, so that all “classes” diverge from one another before any “orders” diverge from each other and so on, but my opinion is that (a) this is not current practice and (b) such a project is both doomed and not very helpful. Above the level of species, it is hard to do it wrong.

However those who would dispose of ranks entirely and say they’re useless are missing a function of ranks: they give hints as to relationships between taxa. I’m not saying it’s an important function – maybe it is, or not – but it is a function.

It is axiomatic that if two taxa are given the same rank in a classification, then they are disjoint. They have no members in common. Nothing can belong to two families or two orders. No reasonable classification has a family contained in a family. Before such a classification is created, the rank of one or the other is changed so that a taxon’s direct child has a different rank from the taxon itself.

Having different ranks does not guarantee that taxa are not disjoint, so the inference only goes in one direction. But when the ranks are the same, disjointness can be inferred just by looking at ranks, with no need to examine the hierarchy.

A second, less important but not useless, function of ranks is that ranks, being ordered, can rule out some possible inclusions. We know that an order can never be subsumed by a genus (within one classification). So we can tell that a given order is not in a given genus just by looking at the ranks, with no need to examine the hierarchy.

So while designated ranks may be of limited utility, they are not useless.

Disjointness and non-subsumption are important pieces of information in any kind of reasoning about collections of taxa. Of course this information can be gleaned by looking at the hierarchy. But having designated ranks gives a shortcut that saves time. We get faster code, and a lighter cognitive load for humans.

From this point of view, ranks are just arbitrary tokens that can be applied within a classification to sets of disjoint taxa at any level. “Family”, “class”, and other tokens no longer mean anything at all except in relation to the tokens given to other taxa. Two taxa that share a token are disjoint. They may not be necessary to the identity (specification) of a taxon, but they can provide a shorthand for relationships to other taxa that bears on the identity of a taxon. If you are wondering what taxon is meant by ‘A’, it might help you to know that A is disjoint from B.

I’d like to say why I was thinking about this – it has to do with RCC-5 and the representation of hierarchies, especially those that include _incertae sedis_ taxa – but this post is already long enough.

ADDED 5/24/2019: I never said I was a scholar about this stuff; this is just a blog. Nico Franz has kindly pointed me to his article (with David Thau) “Biological taxonomy and ontology development: scope and limitations” from 2011 which on page 57 basically says the same thing.

Categories: Uncategorized