I completely agree. From the perspective of deep history, there is no such thing as France, Spain, Italy, Britain, or even Albania, Bulgaria, Georgia, etc. From a musical perspective, we see a clear correlation between singing style and residence in “refuge” areas, such as mountains, islands, forests, a correlation that might well distinguish “Old Europe” from Indoeuropean Europe, to use Gimbutas’ terminology. Future genetic research in Europe (and elsewhere) will need to focus on such distinctions if we are to expect a meaningful reconstruction of the earliest phase of “modern” human history.

In Bauchet-2007 the southern cline was mostly caused by Greek vs. Spanish duality, very clear in the K-means clustering. Basques instead looked weirdly spread around in the PC map because their specificity did not seem to weight in either PC and were therefore subject to minority elements’s randomness, that made them appear intermediate between Spanish and Brits. Basque specifity (very clear) could only be detected in the K-means clustering, and only because they were an important sample (smaller samples are likely to see their possible specificty undetected).

In this paper instead the Greek-Spanish duality seems almost totally supressed, and the Spanish-Italian difference (many in Bauchet) has fully vanished.

I find these differences most suspicious, specially when Bauchet’s study so clearly determined that the main component of Iberians was so distinct from that of Greeks.

PC stats are only somewhat meaningful and are largely dependent on samples and their relative weight. When you sample a large region, only the most widespread components at regional level are likely to show up, while local specificties, even if very dominant locally, won’t.

For that reason I strongly prefer Bayesian K-means clustering, specially when it reaches enough depth. For the case of Europe, so far, Baucher 2007 is the best I have seen so far and it has many shortcomings). But much greater depth has been achieved for instance among Native Americans. I hope to see also a K=16 or greater Bayesian structure analysis of Europe or (better) West Eurasia soon – and I hope it includes some Eastern European and Balcanic samples.

Secondly, the north here shows an E-W cline because of the greater amount of markers used. At the same time, unlike in Bauchet, we’ve got here Northern Greeks as opposed to Greeks, and no Basques or Armenians. Thus the E-W cline in the south looks smaller in this paper.

Btw, here’s a figure I found on the net from the paper that does a better labelling job than Razib…

http://i38.tinypic.com/s3q5na.jpg

In contrast, Baucher et al. (2007) results suggested (in the comparable PC graph) a Mediterranean E-W axis with northern Europeans appearing much more homogenous. Not sure why these differences but may be due to sample bias.

The best thing about this study is that intermediate populations (French, Austrians, Hungarians) are not anymore ignored, what makes the overall result much more clearly clinal.

Nevertheless, I believe sampling should be more concerned not about state boundaries but about actual regional historical differences, not ignoring population sizes either. Hence large heterogenous countries like France, Spain or Britain should probably get regionally representative samples. It is impossible that any single French sample may represent the complex genetic landscape of that state, for instance. Another confusion factor derives from the lack of sampling in Eastern Europe and West Asia.