Upper Paleolithic

Nature this week released two papers about European archaeological sites that come near the end of the Neandertals and beginning of the archaeological transition to Upper Paleolithic industries. Here, I'll devote some attention to the first, by Tom Higham and colleagues [1], which discusses the morphology and dating of the maxilla fragment from Kent's Cavern, England. The paper claims that this is the oldest modern human specimen in Western Europe.

The earliest anatomically modern humans in Europe are thought to have appeared around 43,000–42,000 calendar years before present (43–42 kyr cal BP), by association with Aurignacian sites and lithic assemblages assumed to have been made by modern humans rather than by Neanderthals. However, the actual physical evidence for modern humans is extremely rare, and direct dates reach no farther back than about 41–39 kyr cal BP, leaving a gap. Here we show, using stratigraphic, chronological and archaeological data, that a fragment of human maxilla from the Kent’s Cavern site, UK, dates to the earlier period. The maxilla (KC4), which was excavated in 1927, was initially diagnosed as Upper Palaeolithic modern human1. In 1989, it was directly radiocarbon dated by accelerator mass spectrometry to 36.4–34.7 kyr cal BP. Using a Bayesian analysis of new ultrafiltered bone collagen dates in an ordered stratigraphic sequence at the site, we show that this date is a considerable underestimate. Instead, KC4 dates to 44.2–41.5 kyr cal BP. This makes it older than any other equivalently dated modern human specimen and directly contemporary with the latest European Neanderthals...

One thing you won't see in any of the reporting on the paper: There is no new radiocarbon date for the maxilla.

I must admit, I was completely confused by the paper and had to read the entire thing several times! The first time, I was so busy concentrating on how they obtained their new "date estimate" that I completely missed the one sentence indicating that there is no radiocarbon result.

The supplement gives more details. The radiocarbon dating of faunal specimens from the stratigraphy led the authors to suspect that a 1989 date for the maxilla (30,900 +/- 900 BP) was too young. One woolly rhino and two other bones above the maxilla, over a depth of around a meter, yielded radiocarbon dates around 6000 years older than this. So they went to redate the maxilla, but didn't get enough collagen to obtain a result:

To explore this further, permission was obtained from Torquay Museum to obtain a small sample of dentine from the right P3 of the KC4 specimen for another direct date. The tooth was extracted from the maxilla and carefully sampled at the ORAU so that the external hole could not be seen from the exterior once the tooth had been replaced. Only 89 mg could be drilled due to the small size of the tooth. This produced 0.4% collagen after ultrafiltration pre-treatment, but the total amount extracted was too small for a reliable AMS measurement, so the sample was not dated (Table S2).

So, if they didn't get a radiocarbon result from the maxilla, why are they reporting that this is the earliest modern human in Western Europe?

What they did do: They used the radiocarbon dates on the fauna, and the depth of those faunal specimens in the stratigraphy, to interpolate a date for the maxilla in the absence of radiocarbon information. The Nature paper is simply reporting this interpolation model.

We can look at Figure 3 of the paper to get an abbreviated picture of AMS dates for early Aurignacian human specimens in different parts of Europe. The new Kent's Cavern maxilla date is way out of this distribution.

Figure 3 from Higham et al. [1]. Original caption: " Comparison of direct radiocarbon determinations of AMH bones from European Palaeolithic sites with the KC4 model age. Calibrated using the INTCAL09 curve12. Brackets under the distributions represent the 68.2 and 95.4% probability ranges, respectively. The PDF derived from the Bayesian modelling of KC4 (Model age of the maxilla, in red) is earlier than the original direct date from Kent’s Cavern (OxA-1621) and all others, and overlaps the start of the age range of the earliest European Aurignacian, which is widely accepted as being linked with the earliest AMH. Ultrafiltered collagen radiocarbon dates are indicated with red text; non-ultrafiltered dates are in black. Asterisks denote duplicate dates on the same human bone. The Oase date is a mean of two determinations, one ultrafiltered and one not.

The red distribution is the new model date for the maxilla, way earlier than any other specimen. The gray distribution indicated for Kent's Cavern is the 1989 date, with a calibration model applied to it.

The archaeological association of the maxilla is very weak, as summarized by Higham and colleagues:

The maxilla was found in 1927 at a depth of 10 ft 6 inch (3.23 m) beneath a key ‘granular stalagmite’ used as a datum during excavations undertaken between 1926 and 1941 by the Torquay Natural History Society. Below it were found two blades similar to those discovered in Aurignacian industries, and deeper still were found two blades that resemble those from Initial Upper Palaeolithic industries of the Lincombian–Ranisian–Jerzmanowician complex, which are tentatively associated with Neanderthals.

Such as they are, these associations permit a much later date and do not preclude an earlier one. They are certainly not enough to speak of a date for "Early Aurignacian" on this basis, there is no diagnosis of the industry here.

You can see why I found this so irritating. Here's a paper trying to make a big splash, by establishing the claim in the literature that we have Aurignacian-associated modern human remains earlier at Kent's Cavern than anywhere else in Europe. The reported date estimate is a clear outlier compared to human remains everywhere else. And although there is a radiocarbon estimate, that is ignored (possibly for good reason) in favor of a model that doesn't include it, because radiocarbon gave a date younger than the paper claims, by seven millennia or more.

I'm not saying the authors could have done better with the material they had available. Sometimes we don't get definitive results, and that's expected in paleoanthropology. I just think it's bizarre that Nature would put such press behind a dating paper with no date.

UPDATE (2011-11-07): A couple of people have contacted me, confused by the apparently very ancient dates for other Early Upper Paleolithic sites in the figure. The figure reports calibrated dates, not radiocarbon dates. I have noticed a trend over the last several years to reporting and picturing only calibrated dates instead of the actual radiocarbon determinations. I think this is a very negative development, because it creates confusion between the calibration model and the source of the data. We see how confusing that presentation can be in this paper, where a result that does not come from radiocarbon data is pictured alongside calibrated dates without any distinction between the two.

References

Laurent Excoffier and colleagues' work has investigated how range expansions may have affected human genetic diversity. I've commented on this work several times ("One model, hold the extra parameters", "The Neandertal mtDNA story, 2004 edition, "Surfing and recent selection). They have applied a "geographically explicit" model to questions of human population history, modeling how populations expand and interact in the face of a simulated model of the Old World.

In the past, I've found some things I like in this work, and other points where I disagree with the models' assumptions. Personally, I like to examine analytical models first, because the assumptions are often much more explicit, so we can see more easily how the results follow from them.

This week, a new paper by Mathias Currat and Excoffier in PNAS claims to find evidence for some degree of reproductive incompatibility between Neandertals and modern humans. This is another case where I think the approach is very clever but I disagree with the model's assumptions. I just don't believe that today's distribution of genetic variation can tell us about "reproductive incompatibility" with Neandertals or other archaic people.

Today, if you take a large random sample of people within the continental U.S. and look for a DNA legacy from Precolumbian American people, you will find it in your sample at a level somewhat under 2 percent. This percentage results from the differential growth of European and African-derived peoples during the last 500 years of American history. Whatever else it may be, the current percentage is notevidence for hybrid incompatibility of the world's populations before 500 years ago.

It's not a perfect analogy. Today, Native American ancestry is heterogeneous in the continental U.S., with some people carrying very high fractions. After 30,000 years, such heterogeneity would likely have balanced out. With the Neandertals, we are looking at a much longer history, and different events.

I would contend that the events that have affected today's representation of Neandertal-derived genes were demographically larger than those leading to the European colonization of the Americas. The contraction of the European population during the Last Glacial Maximum, the subsequent movements of Late Upper Paleolithic and settlement of Mesolithic peoples, followed by the introduction of agriculture and waves of population growth and invasions, have partially erased the genetic patterns of the initial Upper Paleolithic. We know that the mtDNA complement of Europe changed markedly both before and after the Neolithic. Today's Europeans are not the people who encountered the Neandertals 35,000 years ago. The genes of those initial Upper Paleolithic people may be almost as rare today as Neandertal genes.

Range expansions and surfing

Nevertheless, I think the analysis in this paper gives us some valuable information about how populations may have interacted at that final stage of population mixture among archaic populations.

A range expansion occurs when a population that is initially limited to some small area begins to expand outward across a larger area. The expansion may include interbreeding with other populations who already occupy those areas, for example, the movement of Neolithic agriculturalists into Europe. Or the range expansion may go into territory where nobody lives, like the initial habitation of the Americas some 14,000 years ago.

Range expansions can distort allele frequencies beyond the pattern expected in a random-mating population. As the population pushes its boundary outward, individuals at the frontier carry with them a slightly skewed sample of the alleles in the population as a whole. Pushing further and further along, this skewed sample gives rise to a founder effect. This phenomenon has been called "allele surfing", by analogy with a spreading wave of population expansion.

When a population expands its range into that of another population, the invaders usually mate with the natives. As the "wave" of migration continues to spread, more and more of the natives' genes are picked up into the expanding population. As a result, you expect to see a gradient of genetic contribution from the original native population, higher and higher as you look farther from the invaders' point of origin.

Currat and Excoffier [1] assume that a group of 50 people originated in Northeast Africa 50,000 years ago and then began to spread throughout the Old World. This population (moderns) expands into the range of another human population (Neandertals) by virtue of a higher carrying capacity: in fact they assume that modern humans existed at four times the density of Neandertal populations. The modern human value is set at 1 person per 10 square kilometers, which is very low compared to ethnographically described hunter-gatherers. The population as a whole is made up of demes that occupy an area 100 km on a side (in some trials, four times as many demes 50 km on a side). The outcome is inevitable: the higher carrying capacity leads modern humans to replace Neandertals, while incorporating some amount of Neandertal ancestry.

Any model is unrealistic to some extent. An unrealistic model generally leads to results that are very different from reality. In modeling, there's a common strategy to deal with this problem: Leave one free parameter and change it until the results fit reality. In this case, the free parameter is migration rate, the probability that an individual will move to an adjoining deme. Currat and Excoffier used values for this parameter that caused their "modern" population to displace Neandertals in Europe over a span of 6000 years. The value that made this dispersal speed was 20 percent per generation for the dispersing modern human population.

I'm a little concerned that a whole literature of geographically explicit population models has emerged in human genetics without any apparent reference to the anthropological literature on human demography. If you know ethnography, a migration rate of 20 percent per generation over 100 km distances seems very high. It's more than double the observed rate of intertribe marriages among precontact Aboriginal Australian people, for example. The value of one person per 10 square kilometers for population density is near the low end ever observed for hunter-gatherers. If it's a stretch to make a model fit with parameters found in known hunter-gatherers, that's when I go back to the drawing board. But then, my philosophy about this is different from most human geneticists. I'm an anthropologist.

Anyway, with these values the result is foreordained: modern humans will replace Neandertals, and fast. What Currat and Excoffier observe in their simulated populations is that the modern humans tend to pick up a larger fraction of Neandertal genes, especially in Europe. How can we explain why our population today has a relatively small fraction of Neandertal genes? In particular, how can we explain why Europeans have no more Neandertal genes than any other population? They conclude that some kind of reproductive incompatibility must have existed.

Where I think the method falls short

I think this paper would be perfectly reasonable if I was willing to assume that the range expansion of modern humans was the last major event in our evolution. If this were true, then echoes of this range expansion would be the most highly visible today — just as astronomers can still find echoes of the Big Bang in the cosmic microwave background.

But I would offer that our genetic diversity today is not the result of a single Big Bang of movement out of Africa. Many population movements of comparable or even larger scale have happened during the last 30,000 years.

The paper presents the hypothesis of reproductive incompatibility as an attempt to solve two problems: First, Chinese, New World peoples, Southeast Asians and Europeans today have approximately the same amount of Neandertal ancestry. Second, the amount of Neandertal ancestry in Europe is only around 2-4 percent. A 6000-year wave of population growth and mixture as modern humans entered Europe might have left more Neandertal genes, and a higher proportion in this Neandertal-rich area of the world than in East Asia.

Here's how I currently see those problems. Europeans today are not the Europeans of the past. They have undergone massive population movements and replacements since the initial Upper Paleolithic people encountered Neandertals. That's not only the result of archaeology, it's also clear from the paleogenetics. If we recognize this subsequent history, then we will find it easy to explain why the rest of the population outside Africa has basically the same small amount of Neandertal ancestry: they received a massive influx of genes from some West Asian population with Neandertal mixture. Europe also got these genes, mostly long after the initial Upper Paleolithic.

So I don't think the present fraction of Neandertal genes tells us anything about sex between Neandertals and humans, except that it happened. Many times. Hooba-hooba.

There is some irony in the timing of this publication, since only last week PNAS published a paper claiming that today's African populations derive some of their DNA from a population fully twice as different from non-Africans as Neandertals were.

I don't fully believe that, either.

References

The science press has its own synchronized cycle, like brain waves, and being in Rome seems to make me into a misfiring neuron. Here it is tomorrow, and there's this story about Neandertals all being dead before modern humans showed up, which for Americans is now yesterday's news. Unless you take the paper NY Times, of course, in which case you probably haven't read it yet.

The occasion for the article is a paper reporting new radiocarbon dates for one of the specimens from Mezmaiskaya, a site in the Russian Caucasus.

The site and the date of the Mez 2 burial

Excavations by Golanova and colleagues have recovered two burials of young children from this site. One of them (Mez 1) has been the subject of much research. Based on some skeletal features and a partial sequence of its genome, the skeleton is a Neandertal child. A sample of one of its ribs was taken for radiocarbon dating, reported in 1999, and yielded a direct AMS date of 29,000 BP. This was out of sync with the other dates from the surrounding level of the site.

Ten years ago, Milford Wolpoff and I suggested that the skeletal features by themselves weren't very convincing, and a recent date (apparently out-of-sync with the surrounding archaeological layer) might signal an intrusive Upper Paleolithic burial [1], despite its Neandertal mtDNA sequence. Now we can look at a large part of the genome of this individual, which is very much like the Vindija Neandertals. In 2005, Skinner and colleagues reported ESR dates from Mezmaiskaya, concluding that layer 3 (including the Mez 1 burial) dates to between 60,000 and 70,000 years ago. By that time, those authors were discussing the inconsistency of the recent 29,000 date for the rib, compared to much older dates (>35,000 BP) for an overlying Mousterian layer. They expected the underlying layer 3 to be much older, and found that to be true of the ESR date estimates.

The second child burial, Mez 2, comes from layer 2 of the site, which is also Mousterian but younger than the first burial. A date around 40,000 years for the Mez 2 infant is basically what was expected six years ago by Skinner and colleagues:

Infant 2 was found in a pit introduced from Layer 2 into Layers 2A and 2B(1). Its’ age therefore is probably about 40 ka. Since the precise surface from which the pit was dug is unknown, this should be considered a maximum age.

That left open the possibility that the burial might be younger.

In the new paper [2], Ron Pinhasi and colleagues report that a sample taken from the infant itself dates to 39,700 +/- 1100 radiocarbon years BP, which calibrates to between 42,960 and 44,600 calendar years BP. The new date confirms that the burial happened relatively soon after the deposition of the surrounding dated bones.

The authors additionally report many other date estimates for faunal materials from the site. These form a pattern in which most are consistent with a relatively narrow range of date estimates, but a few are outliers. One of the important conclusions from the outliers is that contaminated carbon is hard to get out of a sample, even with the advanced ultrafiltration performed by the Oxford lab. The conclusion is narrowly interesting and solid, and it's very important to iron out such inconsistencies -- compare, for example, my 2008 post on the Gorham's Cave chronology.

So what is the big deal?

What does the paper say about the dates of other sites?

Here's where things get interesting. The paper includes this passage in its discussion:

The critical reanalysis of directly dated Neanderthal and AMH fossils from across Eurasia, taking into consideration pretreatment histories and redating results (5), supports our findings in the Caucasus and highlights the lack of reliably dated Neanderthal fossils younger than ∼40 ka cal BP (Fig. 3). Contrary to traditional arguments for up to 10,000 y of coexistence, these data suggest that Neanderthal extinction across Western Eurasia, including the Caucasus, was probably a rapid process, and that coexistence with AMHs, when it occurred, may have been of limited duration.

and this in its abstract:

Our results confirm the lack of reliably dated Neanderthal fossils younger than ∼40 ka cal BP in any other region of Western Eurasia, including the Caucasus.

That last part is a pretty strong statement. No reliably dated Neandertal fossils anywhere after 40,000 years ago?

I thought that was so surprising that I corresponded with the study authors today. One distinct advantage of being in Rome is that I'm synchronized with Europeans, so Tom Higham was able to write back with some of his thoughts. The authors' doubt in the later dates for Neandertal specimens is genuine; their experience is that the newer treatments to remove recent contaminating carbon from samples is eliminating Neandertal dates under around 40,000 years.

A systematic revision of the radiocarbon chronology of late Middle and early Upper Paleolithic Europeans has been underway for several years. This has been an important story, and I've written about it several times (my "dating" category hits most of the posts). I think I once told a journalist that this was the most underreported story in paleoanthropology.

In 2006, Higham and colleagues reported that dates obtained for the Vindija G1 Neandertals, at 29,000 BP, were too young by some 4000 years [3]. That result is listed in the current paper as "doubtful" becuase it did not employ the latest purification strategies. That helps to show that the current paper is "equal opportunity" -- past results from the Oxford Accelerator unit are not immune to doubt. But it is hardly confidence-raising. If we cannot trust radiocarbon determinations made in the last five years, why should anyone submit further samples for testing?

Personally, this was my reaction to the paper: don't grind up any more human bone until the radiocarbon community is unified about sample processing techniques. Let them work it out on the fauna.

The paper lists 15 direct dates on Neandertal specimens younger than 40,000 calendar years BP (some of them multiple samples from single skeletal remains). It lists all 15 of these as doubtful because they do not employ the latest techniques. That is a point emphasized by Higham also (and reflected in several past papers): these date determinatinos are not trustworthy given what we know about sample contamination by recent carbon-14. The Oxford group has put out several papers on this problem. One of the most useful is by Blockley and colleagues [4] because it introduces the device of using the Cantabrian Ignimbrite ash horizon as a marker to compare dates -- dates below the horizon should be consistent, whereas a large sample of actual date estimates include many that are far too young.

At any rate, this is where the story in the linked news article (and others) comes from. University College Cork issued a press release in conjuction with the paper's early edition release in PNAS.

Direct dating of a fossil of a Neanderthal infant suggests that Neanderthals probably died out earlier than previously thought. Researchers have dated a Neanderthal fossil discovered in a significant cave site in Russia in the northern Caucasus, and found it to be 10,000 years older than previous research had suggested. This new evidence throws into doubt the theory that Neanderthals and modern humans interacted for thousands of years. Instead, the researchers believe any co-existence between Neanderthals and modern humans is likely to have been much more restricted, perhaps a few hundred years. It could even mean that in some areas Neanderthals had become extinct before anatomically modern humans moved out of Africa.

This is the lead of the press release. I think that the claim makes up only a minor part of the paper (which is really a results paper about Mezmaiskaya). It is clearly interesting and provocative, but I think the paper's results by themselves do not justify the claim. In the case of Mez 2, a skeleton that the excavators expected to be 40,000 years old, actually turns out to be 40,000 years old. No surprise. There were no incorrect radiocarbon assessments of this specimen, and the apparently wrong assessment of the Mez 1 infant (at 29,000 years old instead of beyond radiocarbon range) has not been corrected here.

Were Neandertals really extinct by 39,000 years ago?

Now, in one sense, the survival of Neandertals after 40,000 years ago is not terribly important. Africans mixed with Neandertals, and as far as we can tell (an issue my lab is addressing now) the mixture is not preferentially within Europe. That argues for a West Asian interaction of the population, and it remains to understand why the ancestors of Europeans did not interact more than other populations. Probably a good hypotheses is that today's Europeans derive most of their ancestry from outside Europe during the last 10,000 years. If the Neandertals did not persist within Europe long during the Upper Paleolithic, that provides another alternative.

But to say that we doubt a particular kind of information about dates is not the same as saying that Neandertals did not exist after 39,000 years ago.

Direct dates on Neandertal bones are far from the only evidence of their persistence in Europe. Dozens of sites are dated by radiocarbon on fauna or charcoal. These dates themselves may be subject to the same critique as applies to the human bone. But there are not 15 of them, there are many, many more.

For example, Gravina and colleagues [5] list 24 AMS dates for Châtelperronian contexts that are 36,000 BP or less. Calibration of dates for this era adds more than 3000 years or more to the calendar years represented by a radiocarbon date, so these are dates likely less than 40,000 years. They may be contaminated by recent carbon (and indeed a few are outliers below 32,000 BP), but if so some of them are remarkably consistent.

Martínez-Moreno and colleagues [6] give a recent review of the Middle-Upper Paleolithic transition in Iberia. They list several sites with late Mousterian industries later than 34,000 radiocarbon years BP, even without counting the contentious examples (like Gorham's Cave) that arguably are later than 30,000 BP.

I would not be happy assuming that every Mousterian site is a Neandertal site, not even in this limited geographic context. There is too much technical overlap, and sometimes small samples of artifacts, to be definitive about such an association. Technology is not biology. Neither would I be willing to assume that late Neandertals are entirely Neandertal -- we see no genetic evidence of African mixture into this population in the Vindija or El Sidron genomes, but these are older than 45,000 years. Who knows what a 35,000-year-old Neandertal in France or Spain (or Croatia) would look like genetically? But only Neandertal remains have thus far been associated with Mousterian and Châtelperronian in France and Iberia. Several sites have stratified Middle to Upper Paleolithic transitions with dates after 40,000 calibrated years BP.

So from the Neandertal point of view, I think this is largely a non-story. There remains substantial question about the pattern of appearance of the post-Neandertal population, as I've extensively discussed here. When we consider the Caucasus, we are still working to understand the timing and mode of the later Neandertals and early Upper Paleolithic people. But there's really no serious challenge to the idea that Neandertals existed in Western Europe after 40,000 years ago.

Or if there is, it'll be out of sync with what most of us think we know.

References