Header image

john hawks weblog

paleoanthropology, genetics and evolution

Photo Credit: Hand remains of Homo naledi. John Hawks CC-BY 2.0

The Guardian reports on a recent cave art discovery in Spain: “Spanish archaeologists discover cave art to rival country’s best”.

Chief site archaeologist Diego Garate said that an estimated 70 drawings were found on ledges 300m (1,000 ft) underground in the Atxurra cave in the northern Basque region. He described the site as being in “the Champions League” of cave art and among the top 10 sites in Europe. The engravings and paintings feature horses, buffalo, goats and deer, dating back 12,500-14,500 years.

If significant cave art remains undiscovered in Spain, imagine what that says about how little we know in the rest of the world.

Neandertal stone circles at Bruniquel Cave

I want to reflect for a moment on the first passage in the recent paper by Jacques Jaubert and colleagues (2016). The paper describes a series of circular structures, made like stone fencerows out of portions of stalagmites, that ancient Neandertals constructed deep within the Bruniquel Cave some 176,000 years ago.

Here are those first two sentences (the rest of the first paragraph introduces the stone structures):

Very little is known about Neanderthal cultures, particularly early ones. Other than lithic implements and exceptional bone tools, very few artefacts have been preserved. While those that do remain include red and black pigments and burial sites, these indications of modernity are extremely sparse and few have been precisely dated, thus greatly limiting our knowledge of these predecessors of modern humans.

This short summary of the state of our knowledge about Neandertal cultures is entirely correct. But at the same time, I think it summarizes only half the problem.

Archaeologists know well that the record of material culture that they can study is biased in various ways, sometimes extreme. They know that material artifacts have multiple uses and purposes within human social systems, and that testing hypotheses about some of the functions of artifacts requires evidence of patterning that can come only from recovering thousands of them across space and time. Even when the archaeological record approaches numbers of many thousands of each type of artifact, as in southwestern France after some 50,000 years ago, the patterns left on the landscape by ancient social systems may blend and mix because the people moved and cultures changed. The cultural patterning that archaeologists find in this later Middle Paleolithic record therefore reflects aspects of human behavior that lie at a larger scale than cultures as we recognize them in the ethnographic present.

Bruniquel circular structures plan view
Figure 1 from Jaubert et al. 2016.

For this reason, archaeologists who are interested in cognitive evolution and the cultural potential of ancient humans by and large cannot rely upon spatial and temporal patterning of artifacts. Or, to the extent that they rely on patterning for evidence about such cultural abilities, they interpret the evidence as entirely negative. The Acheulean is spread across a million years of time and three continents of space, as recognized by the “type” artifact, the handaxe. To some archaeologists, this is powerful negative evidence about the cognitive and cultural abilities of ancient people. After all, if they were clever, wouldn’t they have done something more interesting? Wouldn’t they have cultural patterning on the scale of recent people? The mere existence of some spatial patterning is one of the more powerful arguments for a shift in the Middle Stone Age of Africa toward a more sophisticated array of cultural abilities. It remains unclear, however, to what extent this “change” is merely a result of the increased intensity of investigation and density of evidence from Middle Stone Age sites.

And so, archaeologists of a cognitive bent tend to focus on another category of evidence: artifacts that require a complex series of steps to manufacture, or that are themselves inexplicable without reference to social phenomena. These are the “black swans” of Paleolithic archaeology. They are intrinsically rare in the record, but it hardly matters. The mere fact that a Neandertal might produce birch pitch, with its complex series of manufacturing steps, speaks more about the cognitive abilities of these ancient people than fifty thousand side scrapers.

Going deep into a cave and building stone circles from stalagmites is clearly one such activity. The mere existence of the structures speaks by itself of a level of social and cognitive abilities in the hominins who built them. That is why the report of such a thing is newsworthy, and why it is worthwhile to document so intensively the timeline of their manufacture.

But adding rare things to the numerator of Neandertal abilities is itself a bias, because the denominator is unknown and large.

Across the entire timespan of existence of Neandertals and the branch that gave rise to them, probably fewer than 50,000 of them existed at any time. I would not be much surprised if the true number was much smaller. If the average lifespan of a Neandertal was 20 years, maintaining a population of 50,000 individuals would require around 7 births per day. For the more than half million years this population and its ancestors existed, back past Sima de los Huesos to their common ancestors with Denisovans and African peoples, we can say there were as many as 1.3 billion Neandertals.

Weigh that against how little we know of them, how few pieces of their cultures they left for us. We do not know what the cultural production of an average Neandertal could have been, this is why the denominator is unknown. But it is evident that the fraction represented by the archaeological record is tiny.

The accompanying opinion essay by Marie Soressi is well worth reading, in it she reminds archaeologists of the limits of preservation:

These structures are among the best-preserved constructions known for the whole of the Pleistocene epoch, probably because they were sealed by calcite very soon after they were erected. When the best evidence is found in the best-preserved context, it serves as a reminder for archaeologists of how much we depend on preservation. The fact that some of the art of the period is also often found deep inside caves has been alternatively interpreted as a testimony of the preservation provided by the cave environment3 or as a result of spiritual preoccupations — the underground being a special place4. Perhaps we need to further consider the idea that the fuzziness of the Neanderthal record is due to a lack of preservation.

One might argue that the fact that we only find a tiny number of uniquely informative Neandertal-associated artifact is itself information. Surely, if the Neandertals had been more culturally productive, more modern, they would have left more of a record? Soressi’s essay addresses that argument to some extent, but I want to examine a slightly different point of view. The fact is, it’s hardly just Neandertals. We know very little about the ancient cultures of modern humans.

As recently as the mid-twentieth century, the world’s peoples spoke more than 6000 languages, each representing a culture with some time depth, some antiquity. Yet the vast majority of these languages could be grouped into a much smaller number of families of languages, most of which had their own origins within the last several thousand years. The number of branches leading to extant languages that existed more than 10,000 years ago is today tiny. Ten thousand years ago, the world was full of modern humans. Their languages, their cultures, which must themselves have numbered in the thousands, are mostly today extinct. As we know increasingly from ancient DNA, today’s people derive most of their genetic heritage from only a very small subset of the people who existed 10,000 years ago. That is, early modern humans around the world have left very few cultural traces and many of them have left no genetic traces in today’s peoples. They are gone, almost without a trace.

How surprising is this absence of information from these earlier modern humans? Surely the situation is not so bad as for Neandertals, after all, we have the fairly dense Upper Paleolithic record of Europe, the Maghreb, the Natufian, there are real areas where we have lots of archaeological information from these people between 30,000 and 10,000 years ago. But Southeast Asia and large parts of Africa–home to the majority of the world’s population at the time–have a very sparse archaeological record with many blank spaces. As we are increasingly becoming aware, the first third of human occupation of the Americas is documented by only a handful of archaeological sites. Who knows how many millions of people lived without leaving any record that archaeologists have yet found?

Worldwide, they were not nearly so sparse as Neandertals. Of these early modern humans, by the time they inhabited Australia and entered the Americas, there were tens of millions. If the lifespan of these modern humans averaged 25 years, it would take on the order of 1000 births a day to maintain a population of 10 million humans. From 30,000 years ago to 10,000, a mere 20,000 years of time, some 7.3 billion modern humans would have existed, under this set of assumptions. The true number of course was probably more. This is not the same space as Neandertals; these modern humans lived around much of the world, not just western Eurasia. But without question their population was an order of magnitude larger.

Arguably, our archaeological knowledge of these vast multitudes of people is much greater than what we know so far of Neandertals. As Soressi points out, the archaeological record is biased by time, by the better chance of preserving evidence from these recent time periods, by the kind of artifacts that some of these modern humans made. What is clear is that even for modern humans within the last 30,000 years, vast spans of time and space are complete unknowns.

Again, we do not know the true size of the denominator, but it is large enough to know that the fraction archaeologists see is tiny.

We know so little about the past. As long as this is true, our view of the past will disproportionately be based on the greatest achievements we have found so far. Water crossings imply maritime abilities, technology for sailing and navigating. Regular use of deep underground spaces implies controlled fire, dominance over other fauna that use caves, social systems capable of escape if the light should be lost.

This logic does not apply only to Neandertals, of course. I really like the methodology applied by Jaubert and colleagues, and the paper is basically straightforward. Still I do have a bone to pick with the paper’s review of the use of cave environments by hominins:

Deep karst occupation does not appear to have occurred in Africa in any period, whether the Early or Middle Stone Age, or even the Late Stone Age if we exclude shelters and cave entrances with evidence for human presence in South Africa, Ethiopia and Maghreb (Extended Data Fig. 8). The oldest evidence for the appropriation of this difficult environment is found in Europe, Southeast Asia/Sunda, Wallacea and Australia/Sahul. The accumulation of human bodies by Acheuleans at Sima de los Huesos, Spain (0.35 Ma) is very different from the Bruniquel structures, however. In other examples, the human frequentation of caves is linked to engraving, painting or sculpting activities. These sites are thus younger than 42,000 calibrated years before present and are always associated with Homo s. sapiens. Symbolic, cultural or funerary activities were the main reasons for these cave visits. Until now no evidence has been found for regular Neanderthal incursions into caves, except for a possible case of footprints27, and Neanderthal constructions inside caves, at least at a distance that is no longer exposed to daylight, were totally unknown. Moreover, Upper Palaeolithic constructions in caves are limited to fireplaces, simple hearths, and some rock or speleothem displacements. Even in caves regularly visited since the Aurignacian, constructions are non-existent or anecdotal.

I am very familiar with one African site in which a primitive non-modern hominin species is clearly in a deep cave environment. Homo naledi is there deep in the Rising Star cave with no direct surface access. It’s not quite as far as Bruniquel, but it’s not peanuts, either. That is good evidence for some degree of deep cave access by that species.


Jaubert, J. and many others. 2016. Early Neanderthal constructions deep in Bruniquel Cave in southwestern France. Nature doi:10.1038/nature18291

Soressi, M. 2016. Neanderthals built underground. Nature doi:10.1038/nature18440

The developmental obstetrical dilemma

Alik Huseynov and colleagues have a data-rich paper in the Proceedings of the National Academy of Sciences examining age-related changes in the human pelvis: “Developmental evidence for obstetric adaptation of the human female pelvis”. In the paper, Huseynov and colleagues present a new hypothesis for the evolution of sexual dimorphism in the hominin pelvis: the “Developmental Obstetrical Dilemma” hypothesis.

Here’s the paper’s abstract:

The obstetrical dilemma hypothesis states that the human female pelvis represents a compromise between designs most suitable for childbirth and bipedal locomotion, respectively. This hypothesis has been challenged recently on biomechanical, metabolic, and biocultural grounds. Here we provide evidence for the pelvis’ developmental adaptation to the problem of birthing large-headed/large-bodied babies. We show that the female pelvis reaches its obstetrically most adequate morphology around the time of maximum fertility but later reverts to a mode of development similar to that of males, which significantly reduces the dimensions of the birth canal. These developmental changes are likely mediated by hormonal changes during puberty and menopause, indicating “on-demand” adjustment of pelvic shape to the needs of childbirth.

I really like this paper and the Developmental Obstetrical Dilemma hypothesis. I think the idea makes a great deal of sense. What I see as the most important result of the paper is that the authors find that the female pelvis achieves its maximal inlet dimensions only around age 25-30. As the authors indicate, this is when female fertility is at a maximum. But from an evolutionary point of view, this timing reflects a serious trade-off that has previously not been clearly recognized. Females mature sexually, in many societies marry, and begin having offspring at substantially younger ages than 25.

If a woman is still developing her pelvic inlet at these younger ages, then infants born at those ages face a higher risk of fetal-pelvic disproportion, elevating the chances of a difficult birth and infant mortality, and slightly increasing the mortality risk of young mothers. All this is concordant with conventional wisdom about births to young mothers.

But from an evolutionary point of view, this seems paradoxical. The force of selection is maximal on individuals just at the beginning of their reproductive lives, when they have survived infant and juvenile mortality, and have all their adult reproduction ahead of them. The force of selection on 25-year-old mothers is actually lower than that on 18-year-old mothers. Why do younger women not develop their pelves faster, so that their early offspring will have the lower risk of birth complications enjoyed by children they have in their late twenties?

I don’t think the “pelvic shape on demand” aspect of the Huseynov paper answers this question. But I think the paper’s results point to a reason for the delayed pelvic development. The early component of a woman’s reproductive life, from menarche to maximal pelvic development is essentially a compromise between the fitness advantage of early reproduction and the fitness advantage of a large (and large-brained) offspring. The answer to “why don’t women develop faster” is that if they developed their skeleton faster, they would move first birth even younger than it already is. Natural selection would favor even younger births just to the point that the mortality cost balances the fertility advantage.

When we look at the average age of male first reproduction in traditional societies, it is substantially later than female first reproduction, but much more similar to the time that women gain their maximal pelvic dimensions. From this point of view, the component of a woman’s reproductive lifespan before this time is a roll of the evolutionary dice. Early children contribute very strongly to the growth of populations, so there must have been a strong selective pressure for early reproduction during the origin of modern humans, and probably at many earlier times in our evolutionary history.

The press coverage of this paper has largely focused on one aspect of the analysis; the observation that female pelves continue to change in shape as women age, so that post-menopausal women have slightly more constricted pelvic inlets than younger women. This is a bit unfortunate for two reasons. First, the amount of change in older women is relatively slight. Speculations about the selective value of pelvic changes in older women are probably without much merit, considering that the force of selection on older individuals of both sexes is very small, and the amount of change posited here is so small.

Second, the sample examined here is not a longitudinal sample, it is a cross-sectional sample. The human individuals are a wonderful resource and irreplaceable, and I sincerely hope that CT-scans of these will be made available on an open access basis. But they do not document age-related changes in the same individuals over time, they document the way that dead people of different ages vary. This means that the 80-year-olds do not represent the same set of birth cohorts as the 40-year-olds and the 20-year-olds. This difference is not such a factor when looking at the initial development of the pelvis through childhood and early adulthood, but it does make a big difference when comparing mature and senescent samples. I do not doubt that some age-related changes documented here in this sample are really characteristic of the European population. But older individuals represent a very different early childhood nutritional and disease history than younger individuals in this sample.

In other words, the study does not address, and does not present data to reject, the hypothesis that the 80-year-olds may have already had pelvic morphology at age 40 that was different from the 40-year-olds in the sample. That being said, I do think the comparison with males, which do not exhibit the same pattern of age-related changes, is sufficient to lead us to think that a female-specific pattern exists.

But at any rate, this part of the study is not central to the evolutionary history. Changes to pelvic morphology associated with very elevated ages must be a very recent phenomenon within human populations, and selection associated with them must have been very slight. What the study shows about the development of the pelvis in early adulthood is comparatively profound.


Huseynov A, Zollikofer CPE, Coudyzer W, Gaschoc D, Kellenberger C, Hinzpeter R, Ponce de León MS. 2016. Developmental evidence for obstetric adaptation of the human female pelvis. Proceedings of the National Academy of Sciences, USA (online) doi:10.1073/pnas.1517085113

Mitochondrial evidence of introgression among North American mammoths

Hendrik Poinar and colleagues have a new paper in Frontiers in Ecology and Evolution that reports new mitochondrial genomes from 67 North American mammoth specimens. These include specimens attributed to three mammoth species defined by paleontologists, the Columbian mammoth (Mammuthus columbi), the Jefferson mammoth (M. jeffersoni) and the pygmy mammoth (M. exilis), which they add to pre-existing data from permafrost-preserved woolly mammoths (M. primigenius). This dataset enabled them for the first time to generate a picture of the diversity of this ancient extinct lineage extending into temperate latitudes of North America.

Mammoth tooth detail
Mammoth tooth. Photo credit: James St. John CC-BY

What they found is that the mitochondrial clades exhibit many instances of non-correspondence with mammoth species as recognized by paleontologists. The paper’s discussion of mammoth phylogenetic hypotheses is extensive but a bit cloudy. The main issue is the relationship of Columbian and woolly mammoths. Columbian mammoths were endemic in North America and evolved sometime early in the Pleistocene. Woolly mammoths seem to have evolved relatively recently from the steppe mammoth M. trogontherii, inhabited northern parts of Siberia and Beringia, but reached south of the ice sheets in North America only within the past 125,000 years. After that time, they both coexisted in parts of North America.

Paleontologists have previously hypothesized hybridization between these mammoth species, looking both at skeletal and genetic evidence. Mammoth systematics has relied heavily upon molar morphology, and the number of lamellar plates of the molars is a main diagnostic criterion. Individual mammoths varied in the number of these enamel plates; Columbian mammoths had an average of 18-20 lamellar plates in their molars similar to steppe mammoths, while woolly mammoths tended to have more plates, at a higher density. The woolly mammoth morphology appeared first in Siberia and spread into Europe some 200,000 years ago. Some paleontologists had already suggested that Jefferson’s mammoth is an ecotype within the eastern United States that resulted from hybridization between Columbian and woolly mammoths.

An earlier mitochondrial study by Enk and colleagues (2011) found that two Columbian mammoths had mtDNA types that nest within the variability known for woolly mammoths, this despite the fact that Columbian mammoths were in North America long before the first woolly mammoths appeared in Eurasia. This finding seemed to point to a recent introgression of woolly mammoth mtDNA types into Columbian mammoth populations. What remained was the question of whether this introgression might represent a widespread replacement of Columbian mammoth mtDNA by the woolly mammoth haplogroup, or whether they had just sampled hybrids by chance.

The current study adds dozens of specimens of Columbian and Jefferson mammoths, and these form a fairly clear picture in comparison to the woolly mammoths. Woolly mammoths of Siberia and Beringia belong to three mtDNA clades, which share a common mtDNA ancestor sometime in the Early to Middle Pleistocene. Two of these are quite divergent and relatively rarely found within the sample studied here. One of the clades is very diverse and includes woolly mammoths from Siberia, Beringia, and North America south of the icen sheets. This clade, which the study denotes as “clade 1”, also includes the Columbian and Jefferson mammoths.

The figure from the paper that shows this phylogeny is large and complicated, too much to try to include here – it’s a whole page multi-pane figure with three distinct color schemes and legends describing them. It has geographic, phylogenetic and taxonomic information all mixed together, and while it is a great figure, it’s probably crying out for a better visualization method. So I’m leaving it out of this post.

There is structure within the mtDNA clade that includes both Columbian and woolly mammoths. The Columbian mammoths are all part of a single branch, and the sister of this branch includes mtDNA lineages mostly found in woolly mammoths of eastern North America, along with some Beringian woolly mammoths. Jefferson mammoths are scattered across both these branches. To the extent that the Jefferson mammoths are morphologically intermediate and suggestive of hybridization between woolly and Columbian mammoth populations, their highly diverse mtDNA suggests that they were likely the result of multiple introgression events. In the earlier paper by Eck and colleagues (2011), they compare this situation to forest and savanna African elephants, which have multiple instances of mtDNA introgression as well.

I’m never in favor of interpreting too much on the basis of the single mtDNA phylogeny. The paper agrees that nuclear DNA will be necessary to resolve how introgression and hybridization have really affected the mammoth populations.

That being said, the mtDNA in this case is provocative for more than the apparent mixing represented by the Jefferson mammoths. The mtDNA clade that includes both Columbian and some woolly mammoths seems younger than the origin of Columbian mammoths. The paper goes into substantial detail about why it is difficult to get accurate estimates of the time of mtDNA common ancestors, largely because the “tips” of the branches are separated in part by slightly deleterious mutations that would not persist over very much evolutionary time, so the deeper “roots” of the tree reflect a relatively slower rate of substitutions. But some problem remains even if the age of this clade is somewhat older than the rates suggested in this paper. The paper suggests two possible resolutions. Less likely, the authors propose that the Columbian mammoths had their mitochondrial genomes completely replaced by a woolly-mammoth-derived mtDNA clade sometime after the arrival of woolly mammoths

More likely, the authors propose that mtDNA introgression went the opposite direction, with woolly mammoths taking a North American clade and invading Eurasia with it. They support this by observing that mammoths carrying the clade 1 seem to have included the latest survivors in Eurasia, as if this successful lineage of mammoths had reinvaded and supplanted the earlier, more diverse mammoths belonging to clades 2 and 3. Possibly these woolly mammoths carried adaptive traits with origins in North America that helped them to spread; or maybe they simply exploited an opening left by the decline of woolly mammoths in Eurasia during the later part of the last Ice Age. It is even conceivable that the mtDNA itself was a target of selection, comparison to the phylogeography of nuclear loci would help to settle this.

This scenario is not quite as good a fit to the mtDNA phylogeny; it would require some incomplete lineage sorting to explain why Columbian mammoths are nested within the woolly mammoth clade 1 instead of having the most basal clade 1 subclade. But it fits within the evidence of dynamic replacement of mtDNA within the population of Eurasia.

And it raises interesting possibilities. Woolly mammoths originated from steppe mammoths by reducing body size, shortening the skull, and increasing the number and density of enamel plates in the molars. By the late Middle Pleistocene, they had supplanted steppe mammoths across Eurasia.

In North America, the Columbian mammoth was in many respects the ecological equivalent of the steppe mammoth; so much so that some paleontologists consider them a geographic continuation of the same species lineage. When woolly mammoths re-encountered this species in the Late Pleistocene, they evidently had no reproductive barrier and may have picked up many adaptive traits.

Poinar and colleagues consider what these scenarios mean for taxonomic practice within the mammoths:

Are columbi and primigenius still to be regarded as “good” species if they were capable of introgressing despite a possible million-year difference in their divergence times from trogontherii ancestors? Or is this lengthy difference illusory, because mammoths on both sides of the Bering Strait experienced dynamic population histories of immigrations, contractions, expansions, introgressions and replacements [13; 14; 18], a now well-established finding that throws into question traditional species designations. This point also applies to making assumptions about unidirectional change in morphological attributes, a highly unlikely proposition now that hybridization between supposedly long-separate lineages of North American mammoths has been adequately demonstrated.

From my point of view, all these estimates of timing of population separation and contact seem familiar. The establishment of the Columbian mammoth population in the Early to Middle Pleistocene ran along the same approximate timescale as the differentiation of African and Eurasian archaic human lineages, including the Neandertals and Denisovans. The introduction of woolly mammoths into North America occurred around the same time that modern humans may have been introduced to Asia. The back-migration of North American mtDNA lineages into Eurasian mammoths occurred around the same time that modern humans were dispersing into Australasia and northward into Siberia and Europe.


Poinar H, MacPhee R, Enk J, Devault A, Widga C, Saunders J, Szpak P, Southon J, Rouillard J-M, Shapiro B, Golding B, Zazula G, Froese D, Fisher D. Mammuthus Population Dynamics in Late Pleistocene North America: Divergence, Phylogeography and Introgression. Frontiers in Ecology and Evolution (in press) doi:10.3389/fevo.2016.00042

Enk, J., Devault, A., Debruyne, R., King, C. E., Treangen, T., O’Rourke, D., ... & Poinar, H. (2011). Complete Columbian mammoth mitogenome suggests interbreeding with woolly mammoths. Genome biology, 12(5), R51. doi:10.1186/gb-2011-12-5-r51

Dmanisi and dispersal of Homo from Africa

Discover last April ran a feature article about the finds from Dmanisi. They have made this available online: “The First Humans to Know Winter”. Dmanisi is in some ways a keystone: Earliest site to preserve evidence of fossil humans outside of Africa, earliest site with clear evidence of H. erectus-like cranial and dental morphology, better chronological control than most East African contexts.

The article has an interesting exchange involving quotes from Michael Chazan and Martha Tappan, about the dispersal of early Homo:

Says Chazan: “The problem that keeps you awake, if you think about these things, is that if there was a dispersal event 2 million years ago, before H. erectus, would we see it? If they were using stone tools made of local materials, would we even pick it up? Are we building our models based on things we can’t see?” Dmanisi team member Tappen agrees the site’s fossils are challenging our current understanding of human evolution — but she’s not losing sleep over it.
“As archaeologists, we go with what we have. We make hypotheses and try to test them, and then you dig up something new and go ‘oops.’ And you have to make up a new hypothesis,” says Tappen.
“The Dmanisi individuals are not too different from H. habilis. We should find them dispersing out of Africa 2.5 million years ago,” she explains. “We don’t have that evidence yet, but we have to expect it’s out there.”

That kind of evidence is what some archaeologists claim to have found in Pakistan, with claimed early stone tools from Riwat and the Pabbi Hills; others believe that a pre-erectus habitation of Asia is necessary to explain the morphology of the Homo floresiensis remains from Liang Bua. In light of recent discoveries, I have no reason to think that such a pre-erectus emigration from Africa need have been very much like H. habilis or H. rudolfensis, though.

Jonathan Tennant and colleagues have a new review of the impacts of open access scientific publishing: “The academic, economic and societal impacts of Open Access: an evidence-based review”. The review comments on the costs of traditional subscription publishing:

The question of the current publication cost is difficult and confounded by estimates of the total global publishing costs and revenue. Data provided by Outsell, a consultant in Burlingame, California, suggest that the science-publishing industry generated $9.4 billion in revenue in 2011 and published around 1.8 million English-language articles. This equates to an approximate average revenue per article of $5,000. A white paper produced by the Max Planck Society estimated costs at €3,800–€5,000 per paper through subscription spending, based on a total global spending of €7.6 billion across 1.5–2 million articles per year in total (Schimmer et al., 2014). Other estimates suggest that the total spend on publishing, distribution and access to research is around £25 billion per year, with an additional £34 billion spent on reading those outputs, a sum which equates to around one third of the total annual global spend on research (£175 billion; Research Information Network (2008)).

I had a short Twitter convo the other day with a reader who felt unable to pursue open access publishing because of author fees. I’m no purist; many of my papers are best placed in subscription journals. But fees do not stop me from publishing articles in open access journals. Responsible journals waive fees for authors who do not have institutional or grant support for author fees, and of course many open access journals publish papers without any author fees.


Tennant JP, Waldner F, Jacques DC, Masuzzo P, Collister LB, Hartgerink CHJ. 2016. The academic, economic and societal impacts of Open Access: an evidence-based review. F1000 Research 5:632. doi:10.12688/f1000research.8460.1

Nature has an essay by Alex Csiszar recounting the first episode of peer review by the Royal Society, negotiated between William Whewell and John Lubbock on a paper about orbital motions by George Airy in 1831: “Peer review: Troubled from the start”. It went about as well as you would expect:

Feeling that they had reached an impasse, Lubbock went to the author himself to deliver his suggestions for improvement. Airy was understandably irritated that his manuscript was being subjected to this strange new procedure. “There the paper is,” he wrote to Whewell, “and I am willing to let my credit rest on it.” He had no intention of changing his text. Lubbock threatened to pull out, but ultimately relented and swallowed his criticisms, acknowledging that this was “the first report which the Council have ever made” and trying to see the bigger picture. He thanked Whewell for putting his “shoulder to the wheel” and signed his name to the report.

Seems like the first instance of the referee trying to horn his way in as an author.

Discover magazine did its March, 2016 cover story on the recent hominin discoveries of South Africa, including Rising Star and Malapa, and other important finds. They have now made that article available online: “Rethinking Humanity’s Roots”.

Genetic flatlining of island foxes

Carl Zimmer’s article on “Foxes That Endure Despite a Lack of Genetic Diversity” is interesting and useful:

However the animals arrived on the Channel Islands, they adapted quickly. The oldest island fox fossils date back 7,000 years and show that they were small even then. The Great Shrinking required less than 2,200 years, it seems.
Like other animals, island foxes carry two copies of each gene, inheriting one copy from each parent. In large populations with a lot of genetic variation, there can be many versions of any given gene. An animal may inherit two varying copies of a gene from its parents.
But the scientists discovered virtually no differences in the DNA the foxes had inherited. “We call it genetic flatlining,” Dr. Wayne said.

This is an extreme example of the situation that we consider likely to have generated substantial genetic load in Neandertals. In the foxes, there has been substantial phenotypic change but biologists do not yet know what negative or deleterious effects on the fox population may have resulted from the limited genetic variation. Several of the populations are endangered, and it is hard to know how much of their demographic challenge may be from their intrinsically low genetic variation as opposed to human disturbance to their habitat.

An even better hominin analog may be the Homo floresiensis situation, which was probably nearly as limited in genetic diversity, potentially for a much larger number of generations. Could a hominin population have persisted for thousands of generations in a “genetic flatline” situation?

Lee Berger had been named one of the “Pioneers” in the Time 100 Most Influential People. It’s quite an accomplishment for any anthropologist to make such a list, and it’s great to see the impact of recent fossil hominin discoveries ranking among the most significant news items in the world.

Australopithecus sediba fossils now on MorphoSource

Lee Berger gave the luncheon plenary lecture at the meeting of the American Association of Physical Anthropologists last Saturday, covering the recent discoveries from the Dinaledi Chamber and a broad historical overview of his exploration work.

MH1 skull

At the end of his lecture, he announced that the collection of hominin remains from Malapa are now available freely for download from the MorphoSource site. These remains include parts of two skeletons, designated MH1 and MH2, which belong to the species Australopithecus sediba. Sixty-eight specimens are currently on the site, including the MH1 skull and mandible, the pieces of the MH2 mandible, individual bones of the MH2 hand, and pelvic, vertebral, ribs, and long bones of both skeletons.

These are an incredible collection, and together with the surface scans of the Dinaledi hominin material, they provide a detailed comparison of the anatomy of two species across much of the skeleton.

Much credit belongs to Steven Churchill, who worked long to get these scans uploaded and released in conjunction with Lee’s announcement, as well as the many team members who carried out laser scanning of the material, particularly Scott Williams, Lauren Schroeder, Cody Prang, Ellie McNutt and Daniel Garcia-Martinez.

Short-faced bear convergence and the diversity of American extinct bears

A neat new paper by Kieren Mitchell and colleagues in Biology Letters has an mtDNA phylogeny for some extinct bears of the Americas. The main conclusion is that the giant short-faced bears of North America and South America evolved convergently from smaller ancestors; earlier systematists had generally considered them to be sister taxa.

The first paragraph gives a great review of the past diversity of this group of bears:

The spectacled bear (Tremarctos ornatus) is the only living member of Tremarctinae, a previously diverse group of bears endemic to the Americas. The now-extinct Pleistocene diversity of Tremarctinae comprised the Florida spectacled bear (Tremarctos floridanus), South American short-faced bears (Arctotherium—five species; [1]) and North American short-faced bears (Arctodus—two species; [2]). These species ranged in size from the relatively small Arctotherium wingei (approx. 150 kg; [3]) to the giant short-faced bears Arctodus simus and Arctotherium angustidens, which may have attained body masses exceeding 1000 kg [4,5]. In addition, tremarctine bears displayed a diversity of foraging strategies, ranging from carnivorous/omnivorous (e.g. Arctodus simus, Arctotherium angustidens) to largely herbivorous (e.g. Arctotherium wingei, T. ornatus) [6–9]. The evolution and biogeography of this diverse group of bears is enigmatic, and currently lacks a robust phylogenetic framework.

I had no idea of the past phylogenetic diversity of bears of the Americas, it seems like a really cool story.

Here’s the phylogeny from the paper by Mitchell and colleagues, with all the species included:

American bear phylogeny
Figure 1 from Mitchell et al. 2016. Original caption: Relationships among tremarctine genera resulting from phylogenetic analysis of our mitochondrial genome dataset. Nodes reflect mean age estimates, whereas grey bars reflect 95% highest posterior densities (HPDs). Branch support values (BEAST posterior probability/RAxML bootstrap %) are given for each clade. The approximate temporal range of taxa of interest (see main text) is plotted based on the fossil record, and coloured according to distribution (North America, blue; South America, red). The extant taxon is marked with an asterisk.

The common ancestors of the Arctodus and Arctotherium clades lived in the Early Pliocene. This means that their Late Pleistocene representatives were separated by around 8–10 million years of evolutionary time. Looking at this as a paleoanthropologist, I’m tempted to compare to other similar instances of dietary convergence. For example, the robust australopiths have been argued to represent dietary convergence upon a large-molar and molarized premolar morphology, with small canine and incisor teeth, and large jaw muscles. They last existed during the later part of the Early Pleistocene, and if they were convergent in their anatomical configuration, their common ancestors may have lived in the Middle Pliocene. In this case their divergence may represent some 4 million years of evolutionary time.

Of course, that’s assuming that they are not truly sister taxa. Some evidence points in the direction of convergence, and with such clear demonstrations of convergence in other large mammal omnivore taxa, I don’t think we should rule it out.

The discussion of Mitchell and colleagues’ paper suggests an evolutionary scenario in which Arctotherium reached South America at around the same time as other members of the Pleistocene carnivore guild, including the sabretooths Smilodon and Homotherium and the modern puma and jaguar. The large-bodied Arctotherium first occurs with these predators, with no small-bodied precursor known from the fossil record of South America. They suggest that Arctodus in North America may have evolved large body size with a similar turnover of the carnivore community, in this case after the extinction of large scavengers like Agriotherium and Borophagus and the first introduction of bison into North America.

Obviously these are hypotheses to test further with better fossil samples, but I find them provocative because they do not suggest convergence after vicariance and slow adaptation to changing ecological conditions, the scenario usually described for hominin diversity in the Late Pliocene of Africa. The scenario sketched here for the bears is dynamic. A smaller-bodied Arctodus shifted to larger body size and greater scavenging when the herbivore and carnivore community of North America turned over in the Early Pleistocene. Meanwhile, Arctotherium rapidly entered an empty large scavenger niche as it dispersed along with the large cats into South America. The appearance of Arctotherium is then a parapatric event, and the establishment of the two large-bodied bear lineages in North and South America is due to the chance founding and rapid selection within one species with huge dispersal potential. This is not the isolation of two lineages in relative cul de sacs; it is the effect of rapid adaptation and a relatively small contact zone between the two genera. The weakness of any geographic and ecological barrier is further reinforced by the later dispersal of the extant South American spectacled bear, the relatives of which remained in North America from the Late Pliocene through most of the Pleistocene.


Mitchell KJ, Bray SC, Bover P, Soibelzon L, Schubert BW, Prevosti F, Prieto A, Martin F, Austin JJ, Cooper A. 2016. Ancient mitochondrial DNA reveals convergent evolution of giant short-faced bears (Tremarctinae) in North and South America. Biology Letters doi:10.1098/rsbl.2016.0062

A symposium on the biology and context of Homo naledi

I was in Atlanta with colleagues last week for the annual meeting of the American Association of Physical Anthropologists were last week, in association with the Paleoanthropology Society meeting. I’ll try to give some highlights of these meetings over the next few days.

Teeth from the Dinaledi Chamber
Teeth and mandibles from the Dinaledi Chamber representing different life history stages.

For me, the real event was on Saturday, when 15 members of our team presented detail across much of the skeleton of Homo naledi. This symposium was packed for most of the long morning session, and it was great to see so many friends and colleagues presenting on their research. It was also really useful to have an open half-hour of discussion at the end of the symposium—so many great questions from the audience, and they could be directed to the best person on the team to answer in detail. It was a real scientific conversation, really different from many of the sensationalist debates of past meetings.

Many of the session’s attendees, and a few of the participants, were tweeting during the morning session. Caroline VanSickle has done a great Storify of the tweets from the symposium, covering all the talks and the question and answer period that followed. I was too busy introducing speakers and timing the talks to be tweeting at the same time, so it is great to have this record.

I know the Hannibal meadow muffin story has done the rounds this week, but I find the approach very interesting: Tracing a historical invasion by looking for trace evidence. One of the principal investigators, Chris Allen, has written about the work himself on The Conversation, and it’s worth pointing people to that synopsis: “How ancient horse-dung bacteria is helping our team locate where Hannibal crossed the Alps”.

Using a combination of microbial genetic analysis, environmental chemistry, pollen analysis and various geophysical techniques, we unveiled a mass animal deposition of faecal materials – probably from horses – at a site near the Col de Traversette. The dung, which can be directly dated to around 200BC through carbon isotope analysis (very close to the date on historical records - 218BC), was found at a mire or pond. This is one of the few in the area that could have been used for watering large numbers of animals. The site was originally discovered during geological expeditions to the area, and already fitted descriptions of the terrain – including rockfalls – that Hannibal had to work his way through.

What the revised Liang Bua chronology leaves unanswered

Thomas Sutikna and colleagues report a significant revision to the stratigraphy of Liang Bua cave, which changes the geological age estimates attributed to the fossil and archaeological evidence of Homo floresiensis: “Revised stratigraphy and chronology for Homo floresiensis at Liang Bua in Indonesia”. Earlier work had placed many fossils attributed to H. floresiensis at geological ages younger than 20,000 years, with a last occurrence sometime between 11,000 and 13,000 years ago. Now, the new study shows that all fossil evidence of H. floresiensis is older than 60,000 years ago.

The paper effectively retracts a series of earlier dating results, including the chronologies in key papers by Morwood and colleagues (2004) and Roberts and colleagues (2009). Quoted in a Nature news story by Ewen Callaway, Richard Roberts shared some poignant thoughts about the initial work:

Roberts says that the peculiar geology of Liang Bua would have been hard to notice when the first hobbit bones were found on the final days of the 2003 field season. “Do I think we could have done a better job? Not with what we knew at the time,” he says. “We’re 10 years down the road, and we know a lot more and we’ve excavated a lot more.”

I totally accept that explanation. Cave sites are nearly always complex in some areas of their stratigraphy, and the chronology of a site will change with new information. We have seen redating and revision of stratigraphy in caves again and again in paleoanthropology. This is a normal aspect of the science and even large changes in the chronology have plenty of precedent when we look at the history of the field.

I decided to take a very close look at how the previous papers by Morwood and colleagues (2004) and Richards and colleagues (2009) went wrong, and to what extent other conclusions might be altered by the new chronology. Understanding these mistakes should help us to avoid making similar mistakes in the future. The current paper by Sutikna and colleagues (2016) goes a long way toward reducing inconsistencies in the current chronology of the site. But the new chronology in the paper now contradicts many of the conclusions of earlier papers describing the archaeological and faunal context of Homo floresiensis. Reading the new work carefully and comparing to earlier papers, I don’t understand how to resolve many of the earlier conclusions with the current chronology.

This post is a long one, with several main conclusions:

  1. Although the earlier work by Morwood and colleagues (2004) and Roberts and colleagues (2009) did not recognize a large unconformity with a hiatus between the layer containing LB1 and overlying layers, they did recognize and plot the sloping stratigraphy. The unconformity itself would likely not have led to an incorrect chronology if the previous work had used a better radiocarbon sampling scheme.

  2. Radiocarbon sampling errors were not obvious in these earlier papers because they depicted their sample locations incorrectly in the published site profiles. Morwood et al. (2004) shows that they obtained radiocarbon samples from immediately within the area of the LB1 skeleton, while Roberts et al. (2009) report the same samples in a different location, and new samples at a lower depth than LB1 and within the same stratigraphic layer. According to the new chronology and site profile from Sutikna and colleagues (2016), both those earlier reports incorrectly depicted the stratigraphic positions of their radiocarbon samples.

  3. Both earlier papers reported other dating results that were inconsistent with the radiocarbon chronology, but nonetheless argued that these results were consistent because of weaknesses in the data. As a result, although on the surface the papers seemed to rely upon multiple methods of dating, the entire chronology was actually pinned by the radiocarbon sample locations. It appeared much more solid than it really was.

  4. The new work does not adequately clarify several issues about the Liang Bua chronology, including the last occurrence date of Stegodon in the site, the status of the association of Homo floresiensis with heat-treated stone tools, and the composition of the post-H. floresiensis Pleistocene archaeological assemblage.

  5. Anthropologists routinely overestimate the reliability of geological age estimates, particularly those provided with the initial descriptions of a fossil assemblage. This is a problem for understanding human evolution, and in communicating the science with the public.

Liang Bua cave
Liang Bua cave. Photo credit: Jacques Beaulieu CC-BY-NC.

The details

In the original pair of publications on these fossils in 2004, Morwood and colleagues reported on finds from two distinct areas of the cave: one near the wall and one near the center of the chamber. The LB1 skeleton and a number of other specimens come from the area near the east wall of the cave, and Morwood and colleagues established their position in the chronology by doing radiocarbon testing of charcoal samples, two of which they reported to have come from immediately adjacent to the fossil hominin skeleton. These radiocarbon samples led to the conclusion that the LB1 skeleton was approximately 18,000 years old, and other samples from elsewhere in the excavation showed that some Stegodon remains and archaeological material that Morwood and colleagues attributed to H. floresiensis was as recent as 12,000 years old. A later review and addition of new dates by Roberts and colleagues (2009) did not change the chronology of either area, and introduced additional dates that all confirmed the original findings.

There are several distinct issues of chronology, only some of which are addressed in the new paper:

  1. The chronology of the Sector VII excavation unit that produced the LB1 skeleton and other hominin material, with the initial date for LB1 estimated at 18,000 years old. Later excavation reported by Morwood and colleagues (2009) expanded this excavation area by adding the neighboring Sector XI. The new paper rejects this chronology and provides a new one in which all H. floresiensis fossils are older than 60,000 years.

  2. The oldest occurrence date of skeletal material attributable to Homo floresiensis, which Morwood and colleagues (2004) reported from Sector IV in the center of the front chamber at between 74,000 and 95,000 years old. The new paper changes the relevance of this material but is ultimately unclear about the chronology within Sector IV.

  3. The oldest occurrence of archaeological material in the cave, which Westaway and colleagues (2007) reported at around 190,000 years old. The new paper repeats this and reports no new data to change it.

  4. The latest occurrence of Stegodon and other extinct fauna, which Morwood and colleagues reported at around 12,000 years old. Morwood and colleagues suggested that Homo floresiensis persisted until this time. The new paper is unclear about how the the faunal and archaeological associations of H. floresiensis are changed by the new chronology.

Sector VII situation

Sutikna and colleagues focus most closely upon the excavation area near the east wall of the cave, including Sectors VII and IX reported by Morwood and colleagues (2004; 2009), and later Sectors XIII, XIV, XV and XVI. This area of the deposit includes three different sequences separated by complex surfaces within the excavation areas. The H. floresiensis fossil material comes from a lower sedimentary sequence, which has a sloping and eroded top surface, forming a “pedestal” of sedimentary layers in the excavation area. The new work establishes that the top of this lower sequence has an age older than around 46,000 years. The H. floresiensis fossil remains lie below this date and Sutikna and colleagues identify three datable tuffs and a series of TL and uranium-thorium age determinations that place the fossils older than around 60,000 years.

Liang Bua stratigraphy
Figure 3 of Sutikna et al. 2016. Original caption: "Stratigraphy of excavated Sectors near the eastern wall of the cave. Multiple specimens of Homo floresiensis (LB1, LB4, LB6 and LB8) were recovered previously from sediments now recognized to directly underlie a sequence of five tephras (T1–T5) and interstratified deposits. Together, these remnant deposits form a pedestal, the top of which is dated to ~46 kyr cal. BP (charcoal sample denoted by the red star at ~2 m depth in Sector XXIII). The overlying section—separated here by a white band and dotted lines for emphasis—represents deposits (including three additional tephras, T6–T8) that rest unconformably on the steeply sloping erosional surface of the pedestal. The green stars in Sector VII mark the respective locations (from top to bottom) of the charcoal samples dated to approximately 13.0, 18.5, 18.1, 19.0 and 19.2 kyr cal. BP, which were used erroneously to infer the latest occurrences of H. floresiensis.

Above the pedestal surface, after a substantial period of time including erosion of the pedestal layers, a series of well-stratified, highly sloping layers was deposited. These include several datable layers. This sequence is the original source for the charcoal that was radiocarbon dated by Morwood and colleagues (2004). According to Sutikna and colleagues (2016), these layers have nothing to do with the Homo floresiensis fossils. Above these strata, after a hiatus marked by flowstone, are Holocene deposits from the last 5000 years, including modern human skeletal remains.

After reading this new description of the situation, I expected to go back to the original publication by Morwood and colleagues and find all this stratigraphic detail missing. Instead, I was surprised to discover that they recognized essentially the same things. here is the site profile published by Morwood et al. (2004):

Liang Bua stratigraphy Morwood et al. 2004
Figure 4 of Morwood et al. 2004. Original caption: Stratigraphic section of the Sector VII excavation at Liang Bua, showing the location of the hominin skeleton. Layer key: A, coarse silt; B, silt; C–K, coarse silts; L, tuffaceous silt; M, clay; N (a–d), ‘white’ tuffaceous silts; O, clay and rubble; P, clay; Q, silty clay; R, sandy clay; S, clay with water-rolled volcanic boulders. The circles enclosing the numbers 40 and 42 indicate the locations of luminescence samples LBS7-40 and LBS7-42, respectively, and the squares enclosing the numbers 5, 6 and 7 denote the locations of 14C samples ANUA-27115, ANUA-27116 and ANUA-27117, respectively.

These are the four walls, north, east, south and west, of the Sector VII excavation square. Morwood and colleagues clearly depicted the sloping layers that are identified by Sutikna and colleagues (2016). On the north wall of the sector, those overlying layers are clearly shown at the same depth as the skeleton on the south wall. The unconformity is not there, but the sloping stratigraphy is clearly depicted.

Looking at that site profile, any reader of that paper—including me—would reasonably think that the radiocarbon samples came from amid the skeletal remains of LB1, within the same layer. You can see the “6” and “7” radiocarbon samples pictured there directly within the “skeleton location”, indicated with a dotted line. It doesn’t get much clearer than that. The text refers to the two radiocarbon samples as “associated with the skeleton”. Further, they indicate that they took a luminescence sample (labeled “42”) from immediately adjacent to the skeleton, also within the same layer.

Sutikna and colleagues are saying that this was wrong. The radiocarbon samples were not even taken from within the same layer as the skeleton. If you look at the site profile from the paper from Sutikna and colleagues above, you can see that the sample locations are completely different from the Morwood et al. (2004) profile. The unconformity did not make them different, the original paper apparently just described the locations completely wrong.

The resulting chronological inconsistency might have been caught, if Morwood and colleagues had used the luminescence sample as an independent line of evidence. The luminescence sample from the location labeled “42” in the diagram came back with a TL date of 35±4 kyr, and an IRSL date of 6.8±0.8 kyr. The higher TL date seems inconsistent with it having been taken from the overlying layers that generated the much more recent radiocarbon ages, while the lower IRSL date seems completely out of reason. Neither of them accords well with the new chronology of Sutikna and colleagues, but either might have raised a red flag in the initial publication.

Morwood and colleagues (2004) ended their discussion of the luminescence dating as follows:

The TL and IRSL ages bracket the time of deposition of the hominin-bearing sediments to between 35±4 and 14±2 kyr, which is consistent with the 14C ages centred on 18 kyr.

It was hard for me to understand that conclusion when I read it, but I worked through the text carefully. Instead of seeing these two discrepant dates, with non-overlapping error bars, as inconsistent, they used the physics of the process to argue that they in fact could be used as maximum and minimum age estimates. Therefore, they argued that the TL and IRSL estimates, each different from the radiocarbon date by large amount, were actually an upper and lower bracket that confirmed the radiocarbon age.

Hence an inconsistency became a consistency, and a chronology pinned on only a single piece of information instead appeared to rely upon two independent sources of information. This is a basic error that reinforced the incorrect description of sampling locations.

The work by Morwood et al. (2004) was followed with additional excavation in an adjacent area, Sector XI, and additional dating by multiple methods, all reported by Roberts et al. (2009). Here is the profile diagram for Sectors VII and XI from that paper:

Stratigraphic profile from Roberts et al. 2009
Figure 3D from Roberts et al. (2009).

This figure differs from the earlier profile because of the addition of Sector XI, which doubles the length of the east and west walls of the excavation area. The position of the LB1 skeleton is given by the star.

You can see that the radiocarbon sample locations are reported differently in this profile. In this diagram, the radiocarbon samples are “+” signs labeled with the date in kyr. The original two dates from Morwood et al. (2004) are indicated there as the 18.1 and 18.5 kyr dates. These two samples now appear to lie within the sloping stratigraphy instead of within the area of the skeleton itself. They are at the same depth as LB1 but in an overlying layer. Now two newly-reported radiocarbon samples with calibrated date estimates of 19.1 and 19.2 kyr appear to be in the same level as LB1, but at a lower depth. The circle here is the same luminescence sample reported in Morwood et al. (2004) with its two mutually inconsistent dates.

Roberts and colleagues (2009) did not note this difference in the locations of the radiocarbon samples, now in different stratigraphic positions from those reported in Morwood et al. (2004). However, they did report the date of LB1 differently:

As expected from stratigraphic principles, the calibrated ages gener- ally increase with depth, from a few hundred years in the top 0.3 m of the deposit to almost 20 ka at a depth of 6.5–6.7 m in Sector VII. The 14C chronology therefore straddles the deposits in which LB1 was found, constraining its age to between 19.8 ka and 15.9 ka at the 95% confidence interval (CI) when the 5 samples from between 6.7 m and 4.8 m are taken into consideration. Of these samples, the two closest to LB1 (ANUA-27116 and ANUA-27117) cover a calibrated age range (95% CI) of 19.0–17.1 ka.

They claimed here that the radiocarbon samples actually bracket the LB1 skeleton, and that despite the sloping stratigraphy there is an association of radiocarbon age with depth. They even provide a figure demonstrating this association:

Liang Bua radiocarbon age versus depth
Figure 4 from Roberts et al. 2009. Original caption: "Calibrated 14C age ranges (expressed at the 95% confidence interval in thousands of years before present) for 24 charcoal samples, plotted against the depth (below the modern ground surface) from which the samples were collected; the age of sample ANUA-23607 is not shown. The dashed lines broadly delimit the ages obtained from all four Sectors (III, IV, VII, and XI), which show a general increase of age with depth. In detail, there are differences between sectors, for reasons discussed in the text."

This is as clear a figure as you will ever see showing how easy it is to mislead yourself with statistics.

Roberts et al. (2009) applied this figure to show the regular relationship of calibrated radiocarbon age is with depth in all sectors of the Liang Bua excavation. The figure shows that there is no inversion of dates from the expected increase with depth. Such a figure does lend some confidence in the lack of mixture or turbation of the sediments.

However, here the problem was the sampling scheme. The stratigraphy was complex and sloping. Yet all the radiocarbon dates seem to have been selected within a single vertical transect through the layers. That strategy guaranteed that significant lateral variation in the stratigraphy, including possible unconformities, would be missed.

Roberts and colleagues’ diagram of the Sector VII/XI excavation area recognizes the sloping stratigraphy and underlying pedestal. Their notes on this diagram reflect the differences that are evident between the Sector IV and Sector VII/XI stratigraphic situations. Nonetheless they were still persuaded by the regular relationship of 14C ages and depth within their single vertical transect through the deposit. They claimed that this was strong evidence that LB1 is between 19.8 and 15.9 kyr.

Again, they failed to note the strong inconsistency between different dating methods. Again in this case, the TL dates continued to show much older dates than the radiocarbon, and Roberts and colleagues argued that the TL dates should be considered as “maximum” and the much younger IRSL estimates as “minimum” estimates for the same samples. These thereby defined a range of error from Holocene to 40,000 years ago for some samples, encompassing all the radiocarbon date estimates. Additionally, Roberts and colleagues (2009) had ESR dates on Stegodon enamel, which likewise were systematically older than the radiocarbon chronology, even under the early uptake model which resulted in lower date estimates than the linear uptake model. No Stegodon specimen subjected to direct ESR dating by Roberts and colleagues (2009) had a date anywhere near as young as the 12,000 years that Morwood and colleagues (2004) had proposed as the last occurrence date for Stegodon in the site. Morwood and colleagues (2005) had even claimed that “a well-defined occupation floor” produced H. floresiensis remains together with Stegodon material dating to 18,000 years ago, a claim repeated and supported by Van den Burgh et al. (2009). The ESR determinations should minimally have raised some question about this.

The point is not that the ESR dates were accurate, because from today’s perspective that is likewise questionable. The point is that two sources of evidence cannot be considered independent if one of them is never applied to question or challenge the other.

I don’t think any of these errors were malicious or intentional, but it is important to take account of how papers on geological dating and stratigraphy can go very wrong. These kinds of errors are not rare, as I’ll discuss below; they are common. Understanding them is much like being a crash investigator; you have to understand which parts of the system failed to do better in the future.

I’m satisfied at the description of the situation that Sutikna and colleagues have now provided for Sector VII and the adjacent newer excavation area. The new paper states that the earlier paper by Morwood and colleagues is in error, which for me ends the questions. They provide direct dates on the hominin fossils, and their more intensive sampling has allowed them to construct a much more complete chronology of the deposit and its contents, which has no major inconsistencies.

Sector IV

The new study reports minimal change to the chronology of material from the center of the cave, which Morwood and colleagues (2004) had put between 95,000 and 74,000 years ago. This is the source of the LB2 specimen, for which Sutikna and colleagues provide 234U/230Th age estimates of 71.4 ± 1.1 and 66.7 ± 0.8 kyr. This fossil is within the same range of ages as the Homo floresiensis fossil remains from the area near the east wall of the cave.

However, as I discuss below, there are some serious remaining questions about the Sector IV chronology that Sutikna and colleagues do not address. The sector is a black box in this paper, no profile or additional context is provided, although there are some new date results including the date on the LB2 ulna. From this sector came the majority of evidence of artifacts and Stegodon reported by Morwood and colleagues (2004). Those Stegodon remains generated ESR dates that seem to be inconsistent with the chronology reported in the new paper by Sutikna and colleagues. But I cannot tell from the information provided what exactly is going on with this part of the excavation area.

190,000 years

The oldest evidence of hominin activity in Liang Bua is much older than the H. floresiensis fossil remains. The paper concludes:

The new stratigraphic and chronological evidence for Liang Bua indicates that a pedestal of remnant deposits, dating to more than ~46 kyr cal. BP, has an erosional upper surface that slopes steeply downwards to the north and is unconformably overlain by sediments younger than ~20 kyr cal. BP. All skeletal remains assigned to H. floresiensis are from the pedestal deposits dated to approximately 100–60 kyr ago, while stone artefacts reasonably attributable to this species range from about 190 kyr to 50 kyr in age. Parts of southeast Asia may have been inhabited by Denisovans or other hominins during this period, and modern humans had reached Australia by 50 kyr ago. But whether H. floresiensis survived after this time, or encountered modern humans, Denisovans or other hominin species on Flores or elsewhere, remain open questions that future discoveries may help to answer.

The reference to the stone artifacts as early as 190,000 years ago does not emerge from any new analyses in this paper. This confused me at first, but I followed the authors’ citation to the work of Westaway et al. (2007). That paper reported on an examination of the sedimentary layers exposed in the rear chamber of Liang Bua, as part of the “conglomerate cliff”. They showed that a basal conglomerate layer containing stone artifacts likely was the earliest sedimentary deposit now present in the cave, with a date approximately 190,000 years ago. This is in an area that was not re-evaluated by Sutikna and colleagues.

Aspects that remain unresolved by the new work

The new paper focuses most of its attention on the geological age of the Homo floresiensis remains. While this is an important issue, it is far from the only important issue. The paper does not provide enough information to evaluate several issues that seem to depend closely on the chronology of the site. These include:

  1. The last occurrence dates of Stegodon and other extinct fauna.
  2. Whether any of the H. floresiensis-bearing levels contain evidence of fire or application of heat to artifacts.
  3. How much of the archaeological material formerly “associated” chronologically with H. floresiensis is actually from much later deposits.

If Morwood and colleagues (2004) really were in serious error about the stratigraphy of Liang Bua, all of these issues must be examined, but the new paper is either silent or does not provide a clear answer on them.


Morwood and colleagues (2004) put the last occurrence of Stegodon in the deposit at approximately 13,000-11,000 years ago. The new paper clearly states that this date must be discarded because the associated radiocarbon samples are from the sloping layers that overlie the pedestal of sediment containing Homo floresiensis.

Further, the new paper by Sutikna and colleagues conducted direct U-series dating on seven specimens of Stegodon, finding all of them to exceed 40,000 years ago, and indeed, they suggest that the latest dates here are likely in error because they underlie sediments that have been dated to older dates by other methods:

The Stegodon bone samples (all from Sector XI) span a modelled age range of 80.6 ± 11.3 to 40.5 ± 2.0 kyr, with the youngest minimum age deriving from a bone (U-s-05/LB/XI/51/04) recovered from the same sediments and depth as LB6. Delayed diffusion of uranium into the dense bone matrix of Stegodon may account for the youngest minimum ages appearing more recent than those obtained for the bones of H. floresiensis, the speleothems and the sediments.

The paper later refers to the layers between T2 and T3, which contain Stegodon as well as extinct giant maribou stork fossils, and for which they provide a geological age between 66,000 and 49,000 years ago.

While the paper says all these things, it nowhere gives a clear last occurrence date for Stegodon. The paper provides direct dates only for Stegodon material taken from the new excavation area in Sector XI, and does not revisit material from the original excavation. The new paper seems to leave completely unresolved whether any Stegodon material was recovered at higher levels in the cave.

Morwood et al. (2004) clearly state that Stegodon material was recovered from higher in the deposit, “immediately below the ‘white’ tuffaceous silts derived from volcanic eruptions that coincide with the extinction of this species.” These correspond with the tuffs T7 and T8 in Sutikna et al. (2016), and they provide a date for charcoal just beneath T7 of 12.7 kyr.

Is this the last occurrence date of Stegodon? Or did Morwood and colleagues (2004) also err in their placement of Stegodon material just below T7 and T8? This is no small difference; there’s a vertical difference of more than a meter and up to two meters.

As I read the new paper, the latest date that they provide in association with any Stegodon material is 40,500 years from 234U/230Th, and the paper argues that date is too young, and the Stegodon samples should in fact be older than the speleothem and sediment ages that all exceed 58,000 years ago.

The last occurrence dates of Stegodon and other extinct fauna are pretty central to the question of when modern humans arrived on Flores and what their subsistence activity may have been like. Also, the Stegodon remains were central to Morwood and colleagues’ (2004) argument about a “big game” hunting kit that they associated with H. floresiensis.

On the other hand, the new paper is completely silent about the ESR/U-series dates on Stegodon enamel carried out by Roberts et al. (2009). Most of these come from Sector IV of the cave. Only two specimens are reported with coupled ESR/U-series dates, both of which are vastly older than either the early uptake (EU) or late uptake (LU) models for other specimens. The EU-LU dates from the other teeth each cover a very wide range of dates. Several of the teeth have date ranges between 20,000 and 40,000 years ago, which Roberts and colleagues (2009) saw as unremarkable, considering that the last occurrence was apparently 13,000-11,000 years ago. Now these dates are remarkable. Sutikna and colleagues (2016) refer only to one of these Stegodon ESR dates, with the coupled ESR/U-series date which is much older. They also reiterate that this tooth was found a meter below the LB2 specimen, which now has a direct U-series date of more than 66.7±0.8 kyr. So it looks like the Stegodon should be older than the ESR results from Roberts et al. (2009), although Sutikna and colleagues are silent on this question.

Is 58,000 the last occurrence date of Stegodon in the cave? Is it 40,500? Is it 12,700? Should we use the ESR/U-series dates from Roberts et al. (2009)? I don’t have a clue. I wish the paper had been clear about this important issue.


Morwood and colleagues (2004) put H. floresiensis in direct association with charcoal, which they used for radiocarbon dating. That obviously raised the prospect that H. floresiensis controlled and used fire. This is exactly what Morwood and colleagues (2004) wrote:

Concerning the behavioural context of H. floresiensis, associated small faunal remains include those of fish, frog, snake, tortoise, varanids, birds, rodents and bats. Many are likely to have accumulated through natural processes, but some bones are charred, which is unlikely to have occurred naturally on a bare cave floor.

In a later publication, Morwood and colleagues (2005) reported on burned stones and a “circular arrangement” of burnt stones.

Use of fire by hominins is indicated by charred bone and clusters of reddened and fire-cracked rocks. These include a cluster of three burnt, water-rolled, volcanic pebbles from Spit 84 (840 plusminus 5 cm depth) in Sector VII, and a circular arrangement of five similarly burnt pebbles from Spit 43 (435±5 cm depth) in Sector XI.

Van den Burgh and colleagues (2009) presented a thorough study of the faunal collection to that point, identifying some burned bone elements within the Holocene assemblage, but noting none within the Pleistocene deposits that contained H. floresiensis. That made it appear that Morwood and colleagues (2004) were incorrect to associate burned bone with the H. floresiensis layers.

Brumm and colleagues (2006) further emphasized that Homo floresiensis at Liang Bua co-occurs with heat-treated artifacts. Again, it looked like H. floresiensis must have been intentionally making fires, cooking, and heat-treating stone for tool manufacture.

Moore and colleagues (2009) investigated heat-treating along with other technological elements in the Liang Bua assemblage. They found that heat-treating was more significant in the Holocene levels, with burned and heat-fractured fragments making up 18% of the assemblage. By contrast, these are only 1% of the Pleistocene assemblage. Still, that is 29 fragments of burned or heat-fractured stone in apparent association with H. floresiensis, according to Moore et al. (2009).

Roberts and colleagues (2009) attempted TL dating on one of the burnt pebbles that Morwood et al. (2005) reported from Sector VII, which they suggest as evidence of fire use by H. floresiensis:

The TL-U and IRSL ages of >200 ka for sample LBS7-44, from close to the base of Sector VII, indicate that this burnt pebble was not heated to a sufficiently high temperature to empty the relevant electron traps in the near-surface quartz or K-feldspar grains. At the present time, therefore, evidence for the use of fire by ,em>Homo floresiensis</em> – preserved as charred bone and as clusters of fire-cracked and reddened rocks (Morwood et al., 2005) – is stratigraphically bracketed by TL-B and 14C ages of between ~41 ka and 20 ka.

Are these still in association with H. floresiensis now? Or were they part of the charcoal-bearing deposits that were sampled by Morwood et al. (2004). This is totally unclear from the new chronology. Likewise, it is unclear whether any burned faunal material derives from the same layers as H. floresiensis fossils.

Again, I wish the new paper had been clear about this. Much of the archaeological evidence uncovered by Morwood et al. (2004) came from Sector IV, and Sutikna and colleagues do not change the chronology of this sector substantially. But some of the burned artifacts apparently come from the Sector VII and XI area that has been massively revised. It’s totally unclear from the earlier papers, including Moore and colleagues (2009), who just lumped all early H. floresiensis-associated material into a single stratigraphic unit (Unit 4).

Even if the conclusion is that we cannot be sure about the association of burned remains from the original 2001-2004 excavations, that is important information to share.

Material culture

Sutikna and colleagues conclude with the following statement about the archaeological situation:

All skeletal remains assigned to H. floresiensis are from the pedestal deposits dated to approximately 100–60 kyr ago, while stone artefacts reasonably attributable to this species range from about 190 kyr to 50 kyr in age.

Here is my problem with that statement: Much of the archaeological material “associated” with H. floresiensis has been associated on the basis of a stratigraphic picture that is now clearly wrong. All earlier publications “associated” H. floresiensis with 18,000-year-old charcoal. It may not be so easy to work out from the excavation records which archaeological and faunal material actually belongs to the pedestal sediments, because earlier publications have mostly reported that material in terms of the 10-cm excavation spits instead of stratigraphic layers.

What this means it that some of the archaeological material formerly associated with H. floresiensis on the basis of the incorrect chronology was actually produced by modern humans within the last 20,000 years.

How much? We don’t know.

Sutikna and colleagues do report that the archaeological material from the top of the pedestal deposits, which predate 50,000 years, can be distinguished from the Holocene material on the basis of raw material selection—Holocene modern humans used a much higher proportion of chert. On this difference, they base an argument for the possible persistence of H. floresiensis into these layers between T2 and T3 despite the lack of fossil evidence. But they are silent on the issue of the later Pleistocene deposits—the ones that contain the 18,000-year-old to 11,000-year-old charcoal that were incorrectly associated with H. floresiensis.

The fallible science of geological age estimation

Dates aren’t set in stone, they are part of science, and subject to change as we discover more and better information. Especially with any new discoveries, we should be cautious about assuming anything about their age. The history of the field shows that a geological age estimate is the one thing most likely to turn out wrong in the future!

Ironically, geological age is the one thing that anthropologists seem to assume is the most solid. They often assume that a team’s interpretation of the anatomy of a fossil is just one interpretation, while the geological age stands external to the interpretive process.

When we look at the history of the field, that assumption is clearly wrong. Again and again, important specimens have been published with geological age assessments that were later shown to be off by huge amounts. Rarely have I seen a paper describing a geological age estimate that showed any kind of doubt or hesitancy at all. Most of them read like Morwood and colleagues’ (2004) paper, totally convincing in showing the reasonableness of their approach and the resulting age estimate.

Yet like any other area of science, geological age estimates are sometimes wrong.

Again and again, anthropological teams have relied upon incorrect geological age estimates to inform their interpretation of the anatomy. By relying too heavily on a date, they make mistakes of interpretation that could be avoided by separating the anatomical description from the geological age.

In many cases such over-reliance on geological age has caused them to overemphasize “derived” or “advanced” features of fossil remains, so that they can demonstrate the “first occurrence” of such features. In other cases, anthropologists have overemphasized the “primitive” traits of a fossil to argue for the impossibility of it being an ancestor of other forms at the same geological age or earlier.

The resulting errors of interpretation can be subtle and hard to root out. It is quite true that a later fossil cannot be literally the genealogical ancestor of an earlier one. But this puts rather a lot of stock in an accurate age for the earlier fossil specimen. But the later fossil may well represent a species that existed much earlier in time: first appearance dates are not equivalent to speciation times.

I don’t think we should ignore dates—they are very important for some purposes. But we have to be careful to distinguish the issues of phylogeny, function, and pathology that must be resolved by anatomy alone, and not allow our beliefs about the age of the fossils to bias those analyses.

What about Homo floresiensis?

There seems to be a widespread assumption that radiocarbon dates are the most reliable. The Liang Bua case is just one of many recent cases to show that radiocarbon dates are not infallible. In this case the error emerged from a poor sampling scheme in which samples were selected with insufficient attention to the complexity of the stratigraphic situation.

Look again at the figure from Roberts and colleagues (2009), showing the association of radiocarbon age with depth:

Liang Bua radiocarbon age versus depth

It seemed almost as regular as a natural law, and the LB1 skeleton was clearly bracketed between 15,000 and 20,000 years.

This is such a valuable example showing that anthropologists are very poor judges of which geological data are reliable. When people showed skepticism about the dating of Liang Bua, it was never the radiocarbon chronology they questioned, it was the oldest occurrences of hominin fossils in Sector IV, which seemed so poorly defined by the wide TL and U-series maximum and minimum ages. Yet these ages now appear to be reliable, while the highly accurate-seeming radiocarbon ages turned out to be wrong, because they were poorly sampled with respect to the stratigraphy.

Many other cases have faced similar issues, in which the radiocarbon dates may be correct but their association with fossil hominins or artifacts is questionable. In other recent cases, the problem has been with the radiocarbon methods, as improved ultrafiltration methods have revealed many errors in previous dates obtained with earlier methods. This is true even for some sites dated within the last ten years, as radiocarbon labs presently rely upon different protocols for preparing material.

Most critically, considering that even within the last ten years there have been huge innovations in radiocarbon sample protocols, we should anticipate that further innovation may continue to change methods and dating results in the future.

Science is not really self-correcting; we have to correct it. This means we need to rigorously challenge our own intrinsic biases. As long as we keep testing hypotheses by collecting new data and revisiting old data, we make progress toward identifying errors in previous results. The new understanding of the stratigraphy of Liang Bua is just one step in this process, and we should expect that the geological age of these fossils will continue to be refined. Indeed, the most current result may itself turn out to be wrong, and we’ll need to change ideas again. Stranger things have happened before. Much stranger.

We need to do better informing our colleagues and the public about the process of this science.


Sutikna, T., Tocheri, M. W., Morwood, M. J., Saptomo, E. W., Awe, R. D., Wasisto, S., ... & Storey, M. (2016). Revised stratigraphy and chronology for Homo floresiensis at Liang Bua in Indonesia. Nature. doi:10.1038/nature17179

Morwood, M. J. et al. Archaeology and age of a new hominin from Flores in eastern Indonesia. Nature 431, 1087–1091 (2004)

Morwood, M. J., Sutikna, T., Saptomo, E. W., Hobbs, D. R., & Westaway, K. E. (2009). Preface: research at Liang Bua, Flores, Indonesia. Journal of human evolution, 57(5), 437-449.

Morwood, M. J., Brown, P., Sutikna, T., Saptomo, E. W., Westaway, K. E., Due, R. A., ... & Djubiantono, T. (2005). Further evidence for small-bodied hominins from the Late Pleistocene of Flores, Indonesia. Nature, 437(7061), 1012-1017.

Brown, P. et al. A new small-bodied hominin from the Late Pleistocene of Flores, Indonesia. Nature 431, 1055–1061 (2004)

Moore, M.W., Sutikna, T., Morwood, M.J. and Brumm, A., 2009. Continuities in stone flaking technology at Liang Bua, Flores, Indonesia. Journal of Human Evolution, 57(5), pp.503-526.

Brumm, A., Aziz, F., Van den Bergh, G.D., Morwood, M.J., Moore, M.W., Kurniawan, I., Hobbs, D.R. and Fullagar, R., 2006. Early stone technology on Flores and its implications for Homo floresiensis. Nature, 441(7093), pp.624-628.

Westaway, K. E. et al. Establishing the time of initial human occupation of Liang Bua, western Flores, Indonesia. Quat. Geochronol. 2, 337–343 (2007) doi:10.1016/j.quageo.2006.03.015

Van Den Bergh, G. D., Meijer, H. J. M., Awe, R. D., Morwood, M. J., Szabó, K., van den Hoek Ostende, L. W., ... & Dobney, K. M. (2009). The Liang Bua faunal remains: a 95k. yr. sequence from Flores, East Indonesia. Journal of Human Evolution, 57(5), 527-537.