Pereira et al
BMC Evolutionary Biology
Conclusions: A Bayesian skyline plot placed the main expansion in the time frame of the Late Pleistocene, around 20 ka, and spatial smoothing techniques suggested that the most probable geographic region for this demographic event was to the west of North Africa. A comparison with U6’s European sister clade, U5, revealed a stronger population expansion at around this time in Europe. Also in contrast with U5, a weak signal of a recent population expansion in the last 5,000 years was observed in North Africa, pointing to a moderate impact of the late Neolithic on the local population size of the southern Mediterranean coast.
Update (November 3rd, 2012): The above paper was discussed on the "For what they were ... we are" blog in December of 2010 (link). Maju brings up some very interesting points which I would like to elaborate upon [my comments added in blue]:
Maju: An interesting paragraph, which I'd like to discuss is the following one (from the conclusions section):
The recently revised archaeological dates for the Aterian industry of North Africa emphasize that the makers of this industry do not appear to have left any imprint in the maternal lineages of present-day North Africans. The oldest arrivals amongst extant mtDNAs appear to be the U6 and M1 lineages, which date to 36.6 (24.9; 48.8) and 25.4 (17.9; 33.1) ka respectively . As with U5 in Europe , the arrival time could be older in each case, since the haplogroups appear likely to have arisen within the southern Mediterranean region from haplogroup U and M ancestors, making dating the arrival time very imprecise. Nevertheless, the estimates seem to match best the appearance of the Upper Palaeolithic Dabban industry in Cyrenaïca, as suggested before [15,23].
Maju: I do agree that the antecessor of U6 (and maybe also M1) arrived in North Africa most likely with the Dabban industry and that the chronology estimates confirm that (roughly). However I disagree when they claim that Aterian industry left no genetic legacy. This claim is unfounded because they ignore the L(xM,N) lineages which make up some 25% of North African mtDNA and at least in many cases must be from a time frame older than the Dabban industry, as I discussed in my old blog.
[Marnie: The other possibility is that early U was a trans continental haplogroup in both North Africa and Europe that later expanded in the northwest of Africa and northward into Europe. In that case, U6 would be African and descended from U in Africa. U5 would represent an early Out of Africa expansion across the Strait of Gibraltar or the Strait of Sicily.]
Maju: Another substantial disagreement I have is when they claim that:
Aside from U6, North Africa was also the recipient of European, Near Eastern and sub-Saharan African lineages most of which most likely arrived in the Holocene. Haplogroups H1, H3 and V expanded in Iberia in the Lateglacial/postglacial [11,58-61], and evidently spread into North Africa from Iberia across the Gibraltar Straits, most likely in the early Holocene [62-65].
Maju: This comes from a calibration error, arbitrarily deciding that H1, H3, V and others (H4, H7) as having expanded only after the Last Glacial Maximum. In turn this stems from a fundamental misunderstanding of what archaeologically makes sense: that the only moment of clear contact between SW Europe and North Africa is at the genesis of Oranian culture, which shows clear influence from the peculiar Iberian Gravetto-Solutrean complex (and in turn Iberian Gravetto-Solutrean later shows influence from North African fashions, notably the back-tipped arrow or spear points). While there is lesser connection between Iberia and some scattered coastal spots of North Africa in the early Holocene (Neolithic) with the Cardium Pottery culture this cannot justify almost 30% of the North African mtDNA. There is a more important later "Iberian" influence in North Africa with the Megalithism, however it is hard to explain why these very late hypothetical colonists contributed so much in the maternal genetic pool and so little in the paternal one, not to mention that non-Megalithic Guanches (Canary Island natives) also show more (fossil) Y-DNA R1b and I (17%) than modern mainland North Africans (discussed here).
Maju: The reality of this situation will be better understood when DNA chronologists accept that the timeline of mtDNA H (and V) is older than usually accepted and comparable to that of U, if not even older. In any case my calibration point for these lineages (H1, H3, H4, H7 and V) is precisely the unique trans-Mediterranean connection episode that happened almost without doubt at the genesis of Oranian culture. It is the only safe calibration point, what means that these lineages were formed and consolidated in SW Europe prior to that expansion into North Africa, so they must be at least of Gravettian time-frame (which is c. 22 Ka for Mediterranean Iberia), because Solutrean proper had a very limited presence in this area (two nearby caves in Valencia), even if it gradually modified the strong Gravettian substrate (and was modified by it).
[Marnie: I agree that there is no clear basis for the expansion of H only during the Holocene. The likelihood is that it is an expansion associated with the lower sea level in the Mediterranean during the LGM and up until the early Holocene. In addition to the Strait of Gibraltar, the crossing point could be the Strait of Sicily. The distribution of H1 in North Africa, concentrated on the east side of the Atlas Mountains, favors the Strait of Sicily crossing. A crossing at the time of the LGM would have favored the Strait of Sicily, as the southern Iberian peninsula was likely climatically less favorable than Tunisia and Sicily at this time.]