Wednesday, June 15, 2016

What is Homo naledi, anyway?

There's a new study on the phylogeny of Homo naledi published today in the August issue of Journal of Human Evolution (it's a preprint).  The study is written by Mana Dembo and colleagues.  They compiled a massive matrix of 391 characters (a supermatrix), all from the skull and teeth.  They scored these characters on 22 different hominins and chimps and gorillas.  For H. naledi, they compiled only 123 of those characters from the original bones at Wits (Dembo was on the H. naledi research team).  That's considerably more than the 87 published in the supplemental material of Berger's original description of H. naledi, which I used previously to do my own phylogenetic analysis.

For this analysis, Dembo et al. used Bayesian methods to infer the phylogeny.  I've always been a bit suspicious of Bayesian methods, mostly because of the need for a model for which the probability is known.  That's technically not knowable, but Bayesian methods get around this by drawing model probabilities from a set of "random" models.  So it ends up sort of like a bootstrap in traditional parsimony studies.  What Bayesian methods get you is the ability to test many more models and model parameters than you could with other phylogenetic methods, and that's really a big deal.  In this new paper, Dembo et al. include a fairly nice summary of Bayesian methods for those already familiar with other phylogenetic methods.

They did two sorts of studies.  In the first, they just looked for the best phylogenetic model, which I've copied below (on the left, from their Figure 2).  In this tree, H. naledi ends up basal to a clade that includes modern humans, Neandertals, H. heidelbergensis, and the poorly known H. antecessor (which is thought by some to be ancestral to H. heidelbergensis and by others to BE H. heidelbergensis).  In their second analysis, they specifically tested alternative trees that tested the specific relationships of H. naledi.  They found that H. naledi is definitely a member of Homo, but  they couldn't rule out several alternative trees to the one shown, indicating that the precise phylogenetic position of H. naledi within Homo could not be conclusively determined.

What's interesting is that you can see that in their original analysis as well.  They include the posterior probability values for the clades in their published tree.  Given the way they did their modeling, these probability values end up being like a kind of bootstrap value, so I created a version of their tree that only shows majority-rule clades and leaves the rest of the branches as unresolved polytomies (below, middle tree).  This shows a lot of ambiguity within the Homo + Au. sediba clade, which is very consistent with my own published phylogeny of H. naledi based on 87 characters (Wood 2016, below right).  In both trees, the relationships within Homo are unresolved.

As I explained above, Bayesian phylogenetics allows you to test specific models in ways that traditional phylogenies do not.  Unlike these traditional majority-rule consensus trees, Dembo et al.'s analysis found that Homo relationships were not completely unresolved.  They found that H. naledi is definitely not the sister taxon of African H. erectus, Georgian H. erectus, H. rudolfensis, H. heidelbergensis, and Neandertals.  So the unresolved Homo clade is not really that unresolved:  We can say with some certainty what Homo naledi is not closely related to.

Dembo et al. also report a possible date for Homo naledi of around 900,000 years ago, based on fossil dates and their model of character changes.  I'm not that excited about this date, since the 95% high posterior density interval ran from 2.4 million years ago to the present.  So basically, there's a pretty good chance Homo naledi lived some time in the past 2.4 million years, according to conventional dating, which we already knew.

What does this mean for creationists?  Probably not a whole lot just yet.  It would be nice to affirm that H. naledi is closer to modern humans than H. erectus, since most creationists think that H. erectus is human.  That would boost my own assessment that H. naledi are definitely human descendants of Noah and family.  But Dembo et al.'s results cannot support that kind of specificity.  We just can't say for sure yet.

So what is Homo naledi?  To date, the most comprehensive creationist analysis says they're human.  The new Dembo et al. study shows that they're definitely Homo but probably not Homo erectus.  Other than that, none of us are very certain.

Dembo et al. 2016. The evolutionary relationships and age of Homo naledi: An assessment using dated Bayesian phylogenetic methods.  Journal of Human Evolution 97:17-26.

Feedback? Email me at toddcharleswood [at] gmail [dot] com. If you enjoyed this article, please consider a contribution to Core Academy of Science. Thank you.