May 04, 2011

MCLUST analysis of Mounier et al. (2011) data

I have recently updated my MCLUST analysis of the classical Howells craniometric dataset. Unfortunately, paleoanthropological data are not easy to come by, but the new paper by Mounier et al. (2011) presents scored morphological data on 50 traits for 36 important paleoanthropological specimens.

I began by filling missing values: I used the median of each trait when that was the case. Subsequently, I ran multidimensional scaling on the Euclidean distance matrix, followed by MCLUST on the resulting scalar representation. The number of clusters varied between 4-6, depending on how many dimensions were retained, but was overwhelmingly 4 across dimensions, so I consider 4 groups to be the most stable solution supported by the data.

In the first two dimensions we see a figure quite similar to that for geometric morphometrics in the Mounier et al. paper, with the first dimension contrasting modern humans (right) with archaics (left), and the second dimension contrasting Neandertals (top) with the rest.

The third dimension contrasts Dali (top) from East Asia with the extremely old D2700 from Dmanisi (Georgia).

The MCLUST results are mostly as expected, these are posterior probabilities of assignment of each specimen to each of four clusters; I've labeled the clusters a posteriori based on the skulls that were assigned to them.
The dendrogram of the centroids of the four components matches that presented by Mounier et al.

Discover News covers this paper by claiming that it reinforces the idea that H. heidelbergensis was ancestral to modern humans and Neandertals. I don't think that's borne out by the data, and even the paper does not make that claim, being concerned more with establishing H. h as a valid taxon.

An inspection of the first two MDS dimensions, in addition to the figures of the Mounier et al. paper really makes it clear that both Neandertals and modern humans cannot be very reliably routed to preceding samples. Perhaps there is a hint of that for Steinheim and Sima de los Huesos 5, but none at all for modern humans. Kabwe 1 is clearly not particularly modern or more related to modern humans than any of the other early skulls, so the idea that it could belong to the ancestral lineage that gave rise to Omo must be retired.

Modern humans seem to emerge out of thin air 200-100ky in East/North Africa and West Asia, and Skhul V, Qafzeh 9 (West Asia), Omo II (Ethiopia), Singa (Sudan), LH 18 (Tanzania), and Irhoud 1 (Morocco) all appear to be closely allied to the modern group. Where/how our species emerged, and which are the forms leading to it remain open questions.


eurologist said...

I am convinced the answer lies in South Africa.

The same way Neanderthals developed extremely unique features after climatic isolation ~300,000 years ago, South Africa also became much more isolated. Until then, humans shared genes and features between Central East and North Africa, West Asia, and Europe.

AMH appear out of nowhere in northern Africa and the Levant/West Asia because they migrated there once the climate drastically improved ~130,000 years ago.

Dienekes said...

Fish Hoek and Boskop from Southern Africa lack typical sapiens features despite being of quite recent age, so I doubt modern humans originated there

terryt said...

"I am convinced the answer lies in South Africa".

I'm convinced dienekes was correct at the Heidelbergensis post:

"by homogenizing gene flow between distant populations of a single Homo sapiens species".

eurologist said...

Yeah, but in addition to those you have Border Cave, Klasies River, and Tuinplaas all representing numerous modern features, and going back perhaps as far as ~200,000 years.

Fish Hoek and Boskop are a bit unusual, but so are several of the Levantine finds. Clearly, there are some features (like Schwartz & Tattersall's bipartite brow and inverted-T chin) that seemed to have evolved rather recently (perhaps 100,000 to 80,000 ya) - but at least from the present fossil record, those must be excluded if one wants to identify "archaic modern" humans. We may never find these features present in older fossils.