Botany gets a "C" for species

Nature has a paper in this week about the nature of plant species. It's an interesting claim - according to Loren Reiseberg, Troy Wood and Eric Baack at Indianan University, around 70% of the species identified through traditional taxonomy are, in fact, reproductively isolated lineages, somewhat better than animal species, at around 40%. So much for the vertebrate bias... (although of course animal species also include annelids, arthropods, and molluscs, to mention only a few). Does this mean botanists get a "C", while zoologists get an "F"? Here's the abstract:

Many botanists doubt the existence of plant species, viewing them as arbitrary constructs of the human mind, as opposed to discrete, objective entities that represent reproductively independent lineages or 'units of evolution'. However, the discreteness of plant species and their correspondence with reproductive communities have not been tested quantitatively, allowing zoologists to argue that botanists have been overly influenced by a few 'botanical horror stories', such as dandelions, blackberries and oaks. Here we analyse phenetic and/or crossing relationships in over 400 genera of plants and animals. We show that although discrete phenotypic clusters exist in most genera (> 80%), the correspondence of taxonomic species to these clusters is poor (< 60%) and no different between plants and animals. Lack of congruence is caused by polyploidy, asexual reproduction and over-differentiation by taxonomists, but not by contemporary hybridization. Nonetheless, crossability data indicate that 70% of taxonomic species and 75% of phenotypic clusters in plants correspond to reproductively independent lineages (as measured by postmating isolation), and thus represent biologically real entities. Contrary to conventional wisdom, plant species are more likely than animal species to represent reproductively independent lineages.

One reason why this is an interesting result is that there is what I call a species denialism trend of recent years, which argue that species are merely arbitrary constructs, and should be replaced by terms like "evolutionarily significant group" or "least inclusive taxonomic unit" or "smallest monophyletic taxon", and so forth. Yet it seems there is something there that answers to "species".

My feeling is this: taxonomists have more than a set of definitions and protocols to go by. When they observe phenetic* clusterings in the world, they do so on the basis of a broad range of background knowledge about that general class of organisms. Nobody just picks up a barnacle and says, "hey! this is a species!" unless they already know a hell of a lot about barnacles and biology in general.

There is a formal class of classifier systems in computing known as neural nets (NNs). These are based roughly on the structure of the human cortex, and one of the interesting things about NNs is that they can train on a set of exemplary data, and then make classifications on that basis. Humans are much more complex and adaptable, and for that matter trainable, than NNs in a computer. Consequently, when a suitably informed taxonomist says something is a species, it is very likely she was trained on all the other species in that group known so far, and has a complex classifier system in place. It is no surprise, therefore, that she can identify, using fairly sparse hints, a naturally reproductively independent group.

Even back in the medieval herbals and encyclopedias, species that were identified turn out to mostly be natural (see Stannard's works cited below). In many cases humans can be remarkably accurate at identifying real things with training (which is, I think, the explanation of the oft-cited agreement between Ernst Mayr and the Foré Mountains peoples who both distingusihed the same species of birds, with one exception).

So, why the failure among animals? What's wrong with zoologists? I can't say for sure, but I warrant that it is the nature of the identifying marks or characters used to distinguish them under classical zoology. Most species names and character descriptions date from before phenetic, let alone phylogenetic, classification techniques. Consequently there may be a bias to split when the discriminators are close to the sorts of discriminators that work well in human and mammalian, or even ornithological, cases, and a tendency to lump when you are dealign with velvet worms or mites. Anyone else have any ideas?

Stannard, Jerry (1968), "Medieval reception of classical plant names", in, Actes du XIIe Congrès International d'Histoire des Sciences, 21-31 Août 1968, Paris, Paris.
——— (1979), "Identification of the Plants described by Albertus Magnus' De vegetabilibus lib. VI", Res Publica Litterarum 2:281-318.
——— (1980), "The botany of St. Albert the Great", in Gerbert Meyer and Albert Zimmerman (eds.), Albertus Magnus, Doctor Universalis, Mainz: Matthiàs Grünewald Verlag, 345-372.
——— (1999), Pristina medicamenta: ancient and medieval medical botany. Aldershot: Ashgate.
Stannard, Jerry, Richard Kay, and Katherine E. Stannard (1999), Herbs and herbalism in the Middle Ages and Renaissance. Aldershot ; Brookfield, Vt., USA: Ashgate Variorum.

Hat tip to Nick Matzke for both bringing the paper to my attention and suggesting the post title.

* Phenetic: meaning that the group is defined by a mathematical clustering technique in a Cartesian graph of some chosen variables. Also called numerical taxonomy.