Monday, June 29, 2015

Panbiogeography once more: relatedness between biota, parsimony of dispersal events, and the garden of Eden

For some reason, the first author of Nelson & Ladiges (2001), Gondwana, vicariance biogeography and the New York School revisited, recently sent me that paper with the comment "an item of possible interest". I can only assume that this is because I was anti-convinced that panbiogeography makes sense by the recent contributions of Michael Heads to Australian Systematic Botany.

(Anti-convinced meaning here that not only did they fail to convince me of the virtues of the approach, but the circularity of the first paper's example analysis actually pushed me further towards considering panbiogeography to be unscientific than I was before reading it.)

I always find it interesting to consider arguments challenging my position, but I am not entirely sure how Nelson & Ladiges (2001) is relevant to Heads (2015). The logic of the 2001 paper is rather different from Heads' recent argumentation, and its methodology is rather different from his demonstrative example of a panbiogeographic analysis, so even if I would say "hey, makes sense" after reading it that would still not help the circularity of Heads' analytic approach.

Still, reading it was very insightful; I only now fully appreciate that certain schools of thought appear to be interested not primarily in reconstructing the ancestral areas and movements of lineages (as Heads seemed to be in his recent papers) or in bioregionalisation, but in reconstructing relationships between biota. They try to arrive at something like a phylogenetic tree of biota, where one can say, for example, that the "boreal" region is the sister of the "austral" region.

I find this very remarkable, for several unrelated reasons.

The first is personal: I come from a systematic and taxonomic approach studying individual species, and as thus I have always found it difficult to think in terms of whole vegetation assemblages or biota. I just don't think it makes sense to see biota as individuals or entities that evolve, split or move but instead see them rather as assemblages of lots of individual species that individually and independently from each other do the evolving, lineage-splitting and moving.

But of course one could think in terms of branching relationships of biota - if, that is, they behaved like terminals on a tree. But do they? This is where one might see another case of circular reasoning at work: We first have to assume that only very negligible dispersal takes place between biota before we can even accept such a biota-tree to have any historical meaning. And I think there is very good evidence that lots of movement takes place between biota, meaning that they do not have a phylogenetic relationship but at best varying degrees of similarity.

This 2001 paper calls any inconvenient distribution patterns on a phylogeny "geographic paralogy", in analogy to gene paralogy, that is different genes that only appear to be directly homologous between conspecific individuals but were really derived from each other through a gene duplication event further back in time. The authors then go on to argue that such "geographic paralogy" is uninformative of area relationships and should essentially be ignored.

I think this analogy falls flat, and a better one would be horizontal gene transfer (HGT) across considerable phylogenetic distances, like a retrovirus inserting insect DNA into a plant. Phylogenetic reconstruction of the tree of life works because while HGT does happen from time to time it is sufficiently rare to not obscure overall relationships. And if you know that some conflicting genetic data comes from HGT, then it is fully legitimate to exclude them when you want to find out lineage relationships. Similarly, if dispersal between biota were a similarly rare event, it would in my eyes be equally legitimate to exclude such events when you want to find out biome relationships.

But what this 2001 paper does - throwing out contradictory distribution data as "paralogous", reconstructing area relationships, and then arguing that dispersal doesn't happen - is equivalent to a phylogeneticist throwing out contradictory sequence data resulting from rare HGT, reconstructing species relationships, and then arguing that HGT doesn't happen at all. That conclusion doesn't follow, and in science one can't just throw out inconvenient data with "the latter [relationship speaks]  to who knows what". Well, it might just be evidence for the dispersal events that panbiogeographers reject, that is what it might speak to.

Apart from this, there are a few other issues of interest in the paper. There is the rhetorical question of whether assuming vicariance between regions A and B for (A,(A,B)) would not be more parsimonious than assuming dispersal from A to B because less dispersal events are assumed. The problem is that dispersal events must still have happened, only a panbiogeographer would push them backwards in time, out of sight: The ancestor must have occurred in both A and B, and how did it cover both areas?

Or does a "vicariance biogeographer" always just assume that every ancestor magically poofed into existence across all areas at the same time? That would be extremely convenient, but even if it were realistic in one case it would not work at all levels of the phylogeny; at some point, all the lineages split into smaller areas by vicariance events have to spread out again to be able to undergo the next vicariance event.

The Asteraceae family, for example, occurs worldwide. Did its tribes diverge through vicariance? But then several of its tribes also occur worldwide, so subsequently they must have crossed oceans with ease. Maybe an individual tribe split into sublineages through vicariance? But then several of the genera in those tribes also occur worldwide (e.g. Senecio), so they must have crossed oceans with ease. At some point stuff must have undergone long distance dispersal, especially given that the whole family, much less its tribes, is nowhere near old enough to have spread across land when the supercontinent Pangaea still existed.

Anyway. Point is, vicariance doesn't make for a more parsimonious explanation in terms of dispersal events, it just pushes them backwards.

Finally, the paper has a section that is just one long association fallacy. The authors reject the concept of "centres of origin" for groups of organisms by claiming that it is a creationist idea. Yes, the Bible also has a centre of origin in Eden. So what? The Bible also has diseases in it; yes, we know now that they are not caused by demonic possession, but does that mean that we should reject the existence of diseases?

And what is the alternative to centres of origin: nothing can ever spread? Everything that was ancestral ... somehow ... occurred everywhere already? Again, see above: we can go one step back in time and ask how it got everywhere to start with. Turtles all the way down!

I find it particularly remarkable that this paper ridicules the out of Africa hypothesis of human origins, not least because such a hypothesis does not preclude vicariance within Africa and merely makes the rather unspectacular assumption that humans migrate. Did I write assumption? Sorry, I meant observable fact.

Again one might ask, what is the alternative? Humans originating across the whole planet at the same moment, with our sister lineage, the chimps, subsequently dying out everywhere except in Africa? Because every other scenario must logically assume a centre of origin for our species. Even if the evidence were not completely on the side of out of Africa, how would a scenario of chimps dying out everywhere except Africa be remotely parsimonious or likely?

So yes, this item is of interest, but mostly for the 'tree of biota' approach that I would so far not have expected to be of serious concern to anybody.

No comments:

Post a Comment