Hobbits, Ghosts & Total Evidence

Adrian Currie writes...

I don’t often go for hominid evolution (a mine field I tells ya), but I can’t help but find Homo floresiensis, the so-called ‘hobbits’, fascinating (why will become obvious). A really nice paper has just been published which claims to provide pretty solid evidence about one of the most puzzling of puzzles regarding this puzzling lineage. Here, I want to explore just what such evidence can tell us about the actual history of the hobbits. In particular, I want to emphasize how limited a single line of evidence can be…

*I solemnly swear that I’ll write about Homo floresiensis without making a single Tolkien reference*

*other than that one, I guess*

The stately march of hominid evolution has become more complex and less stately, in a large part due to the tree becoming considerably more bushy: there appear to be many branches and many surprises (and vehement debate about just how bushy things are: I’m especially fond of this iteration). Despite the brimming confusion, the journey from little ape-like critters to upright, big-brained, massively social ones has proven resilient. The story I learnt as a kid - which takes us from what is in effect a short, upright ape, typically represented by Australopithicus, to the still small but much more cognitively fancy Homo habilis (y’know, with the use of tools and bigger brains), to the much taller, athletic Homo erectus (the first out of Africa, or so I thought), to the emergence of the even-bigger-brained Homo sapiens, who also left Africa to become, well, us - is still the story, it just has many more tangents than we expected.

No, I’m not going to use a picture illustrating said march, I’m a bit terrified of saying anything about hominid evolution in the first place—so far as I can tell, pretty much everything is contentious... Ok fine:

I mean, I dunno how monkey-like those early primates were, but you get the idea.

I mean, I dunno how monkey-like those early primates were, but you get the idea.

Anyway, given this kind of progressive narrative, one piece of the story looks truly anomalous. If us hominids were spending our time getting taller, less arboreal, and more brainy, what are a bunch of suspiciously short, tree-adapted and thick folk doing in Indonesia around 70,000 years ago? H. floresiensis are in many ways extremely primitive (this is no derogratory term! It just means that they have a bunch of features which are characteristic of early hominids, and not late hominids).

Me with a life-size reconstruction of  Homo floresiensis , at a wonderful exhibit on hominid evolution at the  Royal Belgian Institute of Natural Sciences.

Me with a life-size reconstruction of Homo floresiensis, at a wonderful exhibit on hominid evolution at the Royal Belgian Institute of Natural Sciences.

The hobbits existed at the same time as ‘behaviorally modern’ H. sapiens. Who knows why they disappeared, but at the very least their presence speaks of the complexity and diversity of hominid evolution. It re-iterates the lesson that we are, in many ways, phylogenetic orphans; the last surviving twig from a once rich tree.

There are two hypotheses about the hobbits’ ancestry worth taking seriously (IMHO). The basic question asks why the lineage has such a suite of primitive traits. One answer identifies Hobbits with H. habilis. On this view, H. floresiensis is so similar to H. habilis because, well, the former are the latter’s descendents, or a descendent of a close relative. If so, the hobits are habiline, and they inherited their hobbity traits.

An, um, not-at-all-to-scale representation fo the first hypothesis. Time moves from left to right, we start with our friend  H. habilis , then the tree splits, with some evolving into  H. erectus  (and eventually those fancy sapiens), and the others eventually become  H. floresiensis.

An, um, not-at-all-to-scale representation fo the first hypothesis. Time moves from left to right, we start with our friend H. habilis, then the tree splits, with some evolving into H. erectus (and eventually those fancy sapiens), and the others eventually become H. floresiensis.

On the other view, the hobbits are not habiline, but erectine: they are the ancestors of the tall, big-brained folk who first left Africa, arriving in South-East Asia around, say, 1.6 million years ago.

Like above, except here we have the erectine hypothesis.

Like above, except here we have the erectine hypothesis.

But hang on, if the hobbits are descendents of H. erectus, what on earth could explain their primitive, habiline features? Well, there is an evolutionary mechanism on hand: island dwarfism. We often see surprisingly small critters on islands, for instance:

Some over-adorable phyletic dwarves. On the left, a pygmy chameleon; the right, a pygmy marmoset. *Squee*

Some over-adorable phyletic dwarves. On the left, a pygmy chameleon; the right, a pygmy marmoset. *Squee*

Phyletic dwarfism is a process whereby normal-sized organisms get smaller over evolutionary time due to the particular features of islands: they are small, isolated, and often have empty niche-space, just waiting for some likely critter to come along (I’m not that quite that naïve about niches, BTW). These features mean that critters arriving on islands are likely to become smaller in order to use up less of the now restricted resources, and to adapt to those new niches. Perhaps the hobbits’ small stature, and other primitive traits could be due to insular dwarfism.

I’m fascinated by the difference between these two hypotheses. Where the habiline hypothesis explains the hobbity traits in terms of their being inherited; the erectine hypothesis explains them as an evolutionary response to changing environments. It’s a lovely illustration of what I think are two distinct explanatory approaches in biology. But here, I want to focus on what evidence we might have in favour of either side.

What do the two hypotheses lean on? Here’s one way of thinking about things. If hobbits are dwarf descendants of H. erectus, then a process of island dwarfism would be capable of producing, not just a little erectus, but an erectus who is oddly similar to that lineage’s habiline ancestors. Could dwarfism do such a thing? Well, there’s a fascinating debate here, which I’m going to look at in my next post. Let’s focus instead on whether hobbits are habiline. If they’re habiline, the hobbity features are not problematic – they’re simply inherited traits. But if the hypothesis is right, our understanding of hominid history is radically transformed.

As I mentioned above, according to the main view about hominid dispersal, we have roughly two radiations out of Africa: that of H. erectus, and that of H. sapiens. Naturally, there is heady debate about the particular details, but that’s the broad picture we have. If, however, the hobbits are the last surviving remnants of the habilines, then that view is wrong: perhaps 2 million years ago, a group of hardy H. habilis (or something similar) left Africa, their descendants surviving up until 70,000 years ago in Indonesia. The only evidence we have of such an early radiation are the hobbits themselves. How likely is such a ‘ghost lineage’? Perhaps quite likely: the chances of any particular animal becoming fossilized, that fossil surviving to the present, and then being found, are vanishingly small. Moreover, some environments are simply not conducive to fossilization. So, we sould expect gaps in the fossil record. However, positing such a ghost lineage is a radical revision of our picture of hominid dispersal: hominids left Africa much earlier, and there were more radiations than we thought.

Here’s where we get to the main point.

As I mentioned, my attention was recently drawn to a new paper in the Journal of Human Evolution which has been getting some traction in the media. In it, Debbie Argue (from my alma mater!) and colleagues provide evidence supporting the habiline hypothesis. The main business of the study is a pretty rich piece of morphological phylogenetics, drawing on characters from the skull, teeth, and so forth. I’m not going to get into the details of the phylogenetics here, but here’s what they take their results to say:

In virtually all analyses, the two best hypotheses for the affinities of H. floresiensis were that it was either (1) the sister to H. habilis alone, or (2) the sister to a “core Homo” branch that includes H. habilis, H. erectus, H. ergaster, and H. sapiens. In the latter arrangement, both H. floresiensis and H. habilis typically appear as successive outgroups to advanced Homo. Thus, the analyses place H. floresiensis as either a close relative to H. habilis, or at a similar evolutionary grade. (17)

Which is to say, the best morphological phylogenetic results we have, point unambiguously at some version of the habiline hypothesis. And the evidence is impressive. In addition to the complex details of morphological phylogenetics, they also point out that there are a bunch of character states which are, as they say, “… unambiguous (i.e., optimization-independent) derived character states” (21). Say what? There are various features of the skull which are not plausibly read as adaptive, so are unlikely to be targeted by selection or effected by it. On this basis, these similarities are surprising if the erectine hypothesis is true. Moreover, several of these traits are unique to H. habilis and H. floresiensis. Further, there are a bunch of diagnostic (and sometimes unique) traits to characteristic of the erectine/sapiens line as well. And indeed, when they produce trees which force floresiensis and erectus to be sister lineages, extremely complex—very unlikely—trees are produced.

Accordingly, there is strong evidence against the hypothesis that H. floresiensis was an island dwarf form of H. erectus. (23)

It’s an impressive piece of work, and one which (I think) should lead us to think more favourably about the habiline hypothesis (to put my cards on the table, the last time I thought about this stuff, for a paper in early 2016, I thought the odds were about even). But, does it mean the thought that hobbits had an erectine origin are dead in the water (as, for instance, the report in the guardian seems to suggest)? I don’t think so.

Consider something which is only mentioned in passing at the conclusion of the paper: “Our phylogenetic results imply a long (.75 Ma) ghost lineage for H. floresiensis, which furthermore likely originated in Africa” (24). The ghost lineage is the main focus of the New Scientist journalistic piece based on it. And indeed as I’ve emphasized, an unknown radiation of hobbit-ancestors leaving Africa 2 million years ago is a fairly dramatic re-writing of the history of the hominid lineage.

But what does this have to do with Argue et al’s results? Well, the very surprisingness of the ghost-lineage tells us something, I think, about the evidence which Argue et al provide for us. Their evidence focuses entirely on phylogenetics: given the way these critters looked, and various assumptions concerning what traits provide the best signals of ancestry, this is the most likely tree. All well and good, but—crucially—this is not our only relevant line of evidence.

As I mentioned above, the evolutionary and developmental mechanisms of dwarfism matter here. If dwarfism might plausibly lead to a kind of ‘reversal’ of some erectine changes, then Argue et al might just be mistaken in taking certain traits as being good signals of ancestry after all.

More interestingly here, are the chances of ghost lineages. Let’s imagine that we knew (don’t mind how) that the probability of a ghost lineage not being detected was vanishingly, vanishingly small. In that case, what are we to make of Argue et al’s evidence? It seems as if we would have very good, countervailing reasons in favour of either hypothesis: the evidence from the paper would lead to confusion, if anything. If, instead, we think that ghost lineages are likely (and moreover doubt that dwarfism could produce those traits) then the evidence should be convincing (indeed, under those conditions we should be already convinced, it seems to me). Okay, so what? There are two things to point out here.

First, and I think less interestingly, the convincing power of the morphological reconstructions do not simply turn on the reconstructions themselves. Rather, they rely on the expectations we have given other considerations—in this case relating to dwarfism and ghost lineages. This is simply to assert a total evidence view about historical reconstruction: even evidence as sophisticated as Argue et al’s phylogenetic work is, doesn’t get to have their voices heard alone.

Second, and I think much more interestingly, in this circumstance some of those other evidence that matters is, well, not easily amenable to quantified, careful examination. How likely is a ghost lineage?. I guess I have a hunch or two, but how would we tell this? For phylogenetics, we have more-or-less well resolved methods for producing more-or-less stable answers to questions about likelihoods (sorta). I’m just not sure what such a thing might look like in this case (although I’d be very interested to see something!). And remember—this question is crucial to how convincing Argue et al’s results are. In this circumstance, it might be that for all of the sophistication of the techniques, they’re still held hostage to evidential considerations which are not.

I don’t think this should lead us to be particularly pessimistic: if such results keep piling up, we might become more confident that the ghost lineage is there, and indeed we could find further evidence of such a thing. However, I think it does have an important lesson about what kinds of conclusions we can draw from single studies: in part the epistemic power a single line of evidence has is held hostage to other features of the hypotheses in question; and indeed sometimes those other features can be frustratingly ephemeral and challenging. We should, then, be cautious of being seduced by evidence simply because it has concrete, systematic characteristics—there could be other, less easily grappled with, considerations which rock the boat.