Tuesday, July 2, 2013

My own take of the species problem

This is the final post in my short series on species concepts. After discussing a selection of synchronous species concepts, with particular focus on the biological one and the Genotypic Cluster Species Concept, and a rather short summary of asynchronous concepts, I will now sketch out my own current thinking on species. Note again that I am not a specialist on speciation or suchlike, and that my conclusions are obviously tentative. But well, as a systematist and taxonomist I cannot avoid dealing with the issue every day, and that includes sometimes problematic controversies even with collaborators, so I have to have an opinion.

1. In theory, I prefer the Biological Species Concept (BSC)

If species are supposed to be (a) a concept with biological reality and (b) a rank that is qualitatively different from other ranks, such as variety or genus, then they have to be reproductive communities. In other words, the BSC is the one that makes species most meaningful. The various concepts that define species as "the smallest group of specimens/populations that has..." are empirical but they do not tell us why we should particularly care about this unit (especially if the character shared by the smallest diagnosable group of specimens is something on the lines of "lower leaf side hairy" or "white patch on the throat of the males"). Reproductive isolation, on the other hand, matters theoretically (see point 4 below), evolutionarily and for conservation purposes.

That being said, when I refer to the BSC I do not mean the absolutist caricature that many people seem to have in mind. I am thinking of reproductive communities, of tokogenetically structured lineages that are significantly isolated from other such lineages, but that does not mean that there has to be zero gene flow or interbreeding between them. The odd introgression event does not matter as long as it is so rare that the character of the two lineages in question is unaffected.

2. In practice, the Genotypic Cluster Species Concept (GCSC) may be the most useful

However, the problem remains that we will very often be unable to test species boundaries under the BSC experimentally. Some organisms cannot be kept in captivity or grown in a garden, some are only known from very few dead specimens, and so on. Thus it is necessary to use proxies, and a very good one is the GCSC. It is empirical and universal, and using it as a proxy for the BSC has a good theoretical justification: If there was a lot of interbreeding between two populations, we would expect there to be a lot of intermediates, and thus no two separate clusters. The only exception would be if the group in question is under disruptive selection, but then we can also reasonably assume that we may be observing incipient speciation - such a situation is known to provide selective pressure towards becoming genetically isolated from each other because accepting genes from the other population reduces the fitness of your own offspring.

Another point is that although the BSC does not apply to asexually reproducing organisms, we may also want to divide them into units that are somewhat comparable to biospecies, and the GCSC can be applied in both situations. It is consequently the concept that I use myself when I do systematics at the species level, either intuitively or quantitatively.

3. Delimiting species along a phylogenetic tree is absurd

Evolution proceeds through the change of allele frequencies in populations; in fact that is pretty much the most popular definition of biological evolution. In other words, evolution is gradual, and speciation is likewise gradual, although it may appear sudden when seen over geological time-scales. Although there are significant gaps between breeding groups existing at the same time and gaps in morphological variation between clusters of organisms living at the same time, there necessarily cannot be any breaks in interbreeding or gaps in morphological variation along the branches of a phylogenetic tree. Applying the BSC or the GCSC asynchronously is consequently impossible.

And indeed it does not really make sense to delimit species along a tree from a cladist perspective. Ancestral species do not suddenly stop somewhere, to be as suddenly succeeded by a descendant species. Instead, they diversify into clades. A clade of two thousand species today = the ancestral species of millions of years ago. It is what the ancestral species has turned into through time.

(An exception to gradual speciaton, by the way, is polyploid speciation where genetic isolation is pretty much instantaneous. But even that would still not confer massive morphological divergence, quite the opposite.)

4. Cladism is incoherent without species, and the acceptance of species logically leads to cladism

Some months ago, a retired colleague who is associated with our institution as a "fellow" argued over lunch that species were the only taxa with a biological reality while all others (genera, families etc) were entirely arbitrary. He scoffed at my "cladist nonsense". The exactly inverse position is held by an American colleague I have mentioned in passing before: He is (like myself) a cladist and thus argues that only monophyletic groups have a biological reality. However, he also argues that species are an entirely arbitrary rank.

My own opinion is that both of their positions are intellectually incoherent. One cannot be a cladist if one believes that there are no species, and if one believes that species have a biological reality then it follows logically that a monophyletic genus also has that reality.

First the latter, because that is perhaps easier to explain. Start from the assumption, as the anti-cladist colleague I mentioned does, that species have a biological reality and can be circumscribed with an objectively and universally demonstrable approach. Now look at monophyletic groups (AKA clades) - what is their definition? Well, they are all the descendants of one ancestral species. It follows logically that if the ancestral species had an objective reality, then so does the clade that is descended from it. Again, the clade is the ancestral species, it is what the ancestral species turned into over time. If you believe in species, you cannot coherently argue that clades do not have an objective reality, and in fact you cannot coherently argue that paraphyletic taxa have the same reality as monophyletic ones.

Now for the other side of the issue, the aforementioned American colleague. This is a bit more complicated but will appear familiar to those who have read previous phylogenetic systematics posts on this blog. The claim is that monophyletic groups are real but species aren't. However, what is a monophyletic group and how do we recognize it? As mentioned in the previous paragraph, it is the group of all descendants of one ancestral species. Working with that definition, it becomes immediately clear that rejecting the reality of species makes cladism incoherent. But okay, we can expand the definition of monophyletic group to "all descendants of one ancestral something". That clears up the first question, but now look at the second one: how do we recognize a clade? Or, to make the problem clearer, what gives us the justification to infer that a group is monophyletic?

In one way or the other, we always need either at least one synapomorphy, that is a character that was fixed in the ancestral "something" of the clade and thus inherited by all its descendants and only its descendants, or we need the group to have a monophyletic shape on a phylogenetic tree, which is of course also inferred from character distributions plus the assumption that the data has a tree-like structure.

If it is the first, we need to assume that the ancestor of our group did not continue to interbreed to any significant degree with organisms outside of the group because then we could not rely on the presumed synapomorphy actually being of diagnostic value. That means that the ancestor needs to have been reproductively isolated. And that means that the ancestor of a clade needs to have been a biological species (or asexual).

If it is the second, we need to assume that our data has a tree-like structure. If the items on the tree are, for example, sexually reproducing individuals or populations (because the "something" that is the ancestor of the clade is an individual or population), then the structure of our data will be tokogenetic and not phylogenetic, in other words it will not be tree-like. That means that the items on our phylogenetic tree need to be reproductively isolated from each other. And that means that the items on our tree need to be biological species (or asexual organisms).

In other words, you can coherently accept species and cladism, or you can coherently reject the reality of both species and clades, but you cannot pick one and leave the other as in a buffet. (In the second case I would argue you are mistaken, but this is about logical consistency here.)


  1. That sounds pretty reasonable to me. Animal species are, I think, generally monophyletic. My impression is that a fair number of plant species are of hybrid origin. I couldn't think of an animal species of hybrid origin. However, I looked around and found a fish species said to be of hybrid origin. The white shiner is a hybrid between the crescent shiner and the common shiner.

    I take it you don't think the Hennig idea of a speciation event marking the extinction of the ancestral species and the origin of two new species is useful. In my work it hasn't mattered one way or the other.

  2. As mentioned earlier, in my opinion the term monophyletic cannot actually be made to apply to sexually reproducing species, it would be something like a category error. If the phrase is meant to indicate that the all alleles of every individual gene in a species are monophyletic relative to related species then no, there are papers showing that incomplete lineage sorting exists in animals just as in plants.

    Animals may have it easier to select their mates and thus avoid hybridization. But even in plants I suspect that the evidence for hybrid speciation is probably being overestimated. In how many cases has somebody observed incomplete lineage sorting and shouted "evidence for hybridization!" because they were quite simply unaware of the former possibility or did not have a test on their hands to distinguish the two scenarios?

    I also think Hennig's internodon species concept is entirely irrelevant for classificatory practice, not least because most of us work with extant species only.