How not to convince me that panbiogeography is science

A few days ago, the newest issue of Australian Systematic Botany appeared, and I was rather surprised by the contents, for two of the four papers of the issue were opinion pieces by panbiogeographer Michael Heads. One of them is even, to the degree that it discusses anything specific at all, about ratites. Why would a journal of systematic botany accept a zoological paper from a fringe school of biogeography? And why would a panbiogeographer submit his manuscripts to a regional taxonomic journal as opposed to an international one focusing on biogeography? One wonders.

I have written about panbiogeography before, and my perception was and still is that its proponents are mostly characterised by an irrational hostility to long distance dispersal. In my second post I discussed an example of Heads himself trying to force a biogeographic story into vicariance although that would mean that the plant family in question would have had to originate before the evolution of multi-cellular life.

But as in the case of 'evolutionary' systematics I remain interested in the arguments of the other side, and thus I decided to read the two papers. I have started with and will make this post about Heads' Panbiogeography, its critics, and the case of the ratite birds because it promises to provide an example of the panbiogeographic method in action.

After a short introduction, Heads first sets out to rebut what he sees as four major criticisms of panbiogeograpy: (1) that it is creationist, (2) that it does not accept that plants and animals disperse, (3) that it is disingenuous, and (4) that it ignores or misrepresents key evidence.

I'd say that points 1, 3 and 4 are essentially identical, but interestingly Heads appears to take the charge of creationism rather literally; the title of his subsection is really 'panbiogeography is creationist', but most reasonably complete quotes that he provides of his opponents make clear that they merely compare that school of biogeography to creationists because they perceive a similar unwillingness to accept contrary evidence in both cases. At any rate, I will happily grant that it isn't creationist and move on.

A similarly literalist approach characterises the discussion of the claim that panbiogeography does not accept dispersal. Yes we do! Writes Heads, and cites the fact that panbiogeographers explain contemporary sympatry (co-occurrence of different species) as the result of dispersal. True, that is the logical consequence of accepting vicariance as the only speciation mechanism, but that is clearly not what the critics are talking about. They criticise the school for not accepting founder events as a speciation mechanism and in particular for ruling out the possibility of long distance dispersal.

The second half of that section turns, in my opinion, particularly bizarre. By quoting several evolutionary biologists who argued that speciation is probably more often the result of founder effects than of vicariance, Heads manages to insinuate that biogeography traditionally ignored the possibility of vicariance pretty much completely, and that it is only thanks to the tireless efforts of panbiogeographers that it is receiving more attention now. There are several points that occur to me here. The first, and most obvious one, is that one needs to distinguish between evolutionary biologists studying speciation on the one side and biogeographers on the other. There is some overlap, but they are basically two different research areas.

The second is that Heads presents no actual arguments in favour of the idea that vicariance is the most likely speciation mechanism, he just takes that as a given. From a genetic perspective it seems fairly clear that it is much easier to achieve speciation through a small bottle-necked founder population than through a geological event that produces two very large daughter populations. What is more, how often do the kind of events occur that panbiogeographers accept as potential reasons for vicariance? As demonstrated in the present paper, they are pretty much limited to the uplift of mountain chains and plate tectonics. Does that kind of separation of populations happen often enough to explain the millions of species that exist today and the many millions more that have existed in the past? Surely not.

Finally, the ludicrous idea that everybody except panbiogeographers is against vicariance and had to be convinced to take it into account looks like a severe case of projection. In reality, mainstream biogeographers have always accepted vicariance as one of several options, and it is the panbiogeographers who reject some options out of hand. (All except one really.)

And let us look at actual mainstream biogeographic practice, shall we? To infer ancestral areas of distribution, several formal algorithms have been developed over the years. The oldest one is known as DIVA. What does that stand for? DIspersal and Vicariance Analysis. Not only is vicariance right there in the name, it is well known that the method mercilessly favours vicariance over dispersal. Unless you restrict the number of ancestral areas to something realistic (my suggestion would be the number of areas that the most widely distributed extant species occupies) it will simply add all the descendants' areas up and arrive at the solution favoured by panbiogeography: the ancestor occurred everywhere!

The two younger methods that are now available are Likelihood Analysis of Geographic RANGEs (LAGRANGE) and a very new Bayesian method. I have less experience with their behaviour, but Nick Matzke recently spent a good part of his Ph.D. developing variants of those approaches that allow the possibility of founder events - because so far they didn't. In summary, the traditional formal methods used in historical biogeography heavily favoured vicariance and completely ignored founder events and long distance dispersal. Methodology started out as essentially panbiogeographic, only swapping the intuitive track analysis for formalised algorithms in an explicitly phylogenetic context, and only recently have its practitioners started to include the processes that panbiogeograpers reject. This is pretty much the opposite of Heads' story.

The section 'panbiogeography is disingenuous' claims that the sea floor of the pacific ocean has dropped thousands of metres in recent times. If this were true, it would provide a convenient mechanism for the long term persistence of the species that are now in Hawaii; they could have existed on different islands that are now under water instead of having recently undergone long distance dispersal from Asia and America.

A bit too convenient, one might say. The pacific is huge - where were those massive amounts of water in the meantime? Wouldn't most of the continents have been under the ocean? It seems fairly unbelievable, but I will readily admit that I am out of my area of expertise here and leave that up to the geologists, not least because even if Heads were correct this would still not address any other cases of recent long distance dispersal, e.g. from Australia to New Zealand (e.g., Craspedia, Ozothamnus) or between Africa and South America (e.g., Rhipsalis, certain Acanthaceae). And remember, nobody claims that there is no vicariance; the mainstream position is merely that long distance dispersal is not impossible.

This is followed by the 'ignores evidence' section which basically only says that panbiogeography does too use all evidence... but the specifics are all about geological evidence, which at best may show that vicariance was one of several possibilities, and not, for example, about time-calibrated phylogenetic analyses or the empirical evidence for contemporary long distance dispersal.

There follows a long section on the aforementioned time-calibrated phylogenies. As indicated above, I have written about this issue in more detail elsewhere, so not much needs to be said. The present paper conveniently ignores the contribution by several mainstream scientists showing that the sunflower family would have to be more than one billion years old if the distribution of Abrotanella, which Heads built his criticism of this approach on, were to be explained through vicariance.

After some more discussions of Pacific geology which are, once more, beyond my ability to assess, we arrive finally at the example case of the ratites. At least as presented by Heads in this paper, the panbiogeographic approach in practice is like an intuitive, pencil and paper DIVA (see above). Major points of interest are that it starts from the assumption of a widespread ancestor and then merely searches for geographic events that have occurred in the same order as the lineage splits on the phylogeny. The first does, of course, guarantee that the answer will only ever be vicariance, and the second could be taken as rationalisation; if the time depth of the tree can be adjusted however you like (because time calibration cannot be trusted, natch), then one will always find some event that can be interpreted to have been responsible. (And of course one can ignore any that don't fit; who says that every barrier must lead to speciation?)

What is most interesting to me here is not so much what this example case demonstrates but what it doesn't. It could well be said that you are not doing science unless you ask yourself: how would I know that I am wrong? What would therefore have been much more helpful, if the present paper is intended to convince people of the bona fides and respectability of panbiogeography, is a case in which long distance dispersal is accepted even by that school. How would their method show that there really is no vicariance event? This would immediately demonstrate the criticism of it dogmatically rejecting such a possibility to be misguided.

Alas, the ratite example in the present paper merely shows how one can arrive at a panbiogeographer's preferred conclusion. And in fact it shows that any panbiogeographic analysis can only ever do so. As Heads himself openly admits: "The panbiogeographic approach, as illustrated here, explains allopatry by vicariance and overlap by dispersal". In other words: it dogmatically excludes the possibility of allopatry being explained by dispersal (and of overlap being explained by sympatric speciation).

If that is the case, how can it be science?