Thursday 27 April 2017

Extraordinary claims require extraordinary evidence

The history of claims for the earliest human presence in the Americas is littered with extraordinary claims, many of which have not stood up to scrutiny. In many cases it has been the dating rather than the human associations of the site that has failed to withstand other scientists' examination. This week Nature has published a paper by Holen and coworkers on “A 130,000-year-old archaeological site in southern California, USA”. Of great interest to me is that the dating is primarily based on “230Th/U radiometric analysis of multiple bone specimens using diffusion–adsorption–decay dating models”. The first development of this approach to dating bone was in my doctoral thesis and it was published in my first scientific paper. But I've also published quite a few papers on re-evaluating the claimed dates for Palaeolithic sites and the minimum level of information that needs to be published for readers to be able to assure themselves of the quality of dates. So, how does this look?

The uranium-series dates


The authors are to be congratulated in making the full set of radiometric data available in the supplementary information, and making it available in machine-readable from from the USGS ScienceBase. The main text reports a series of checks on the quality of the analytical chemistry, and these are all passed with flying colours. The key to understanding the quality of the uranium-series dates is in Extended Data Figure 10 (this is free to access even though the main paper is not). The diffusion-adsorption model makes a series of assumptions and predictions that are not directly involved in the date calculation using the updated approach used here. The things to watch out for are:
  1. the concentrations of uranium through the bone should be U-shaped, high at the edges and lower in the middle. In the top panel of part (a) of the figure this appears to be broadly the case for all three bones dated, but CM-292 has a decrease towards the right of the plot, which might indicate inhomogeneity in the bone or leaching of uranium (which the model assumes does not happen).
  2. the distribution of apparent uranium-series ages calculated without making any correction for uranium uptake, as shown in the lower panel of part (a) of the figure should also be U-shaped. For this test the data is noisier but follows the expected trends. Here CM-292 is the best of the three samples. The iDAD software used to calculated ages can accommodate scatter like this – it makes the final date less certain.
  3. comparison of the predicted isotope ratios from the best fit model to the actual data, as shown in Panel (c) in the second and third columns. CM-292 clearly has the best fit, and fairly smoothly varying observations, while the other two samples have more jagged distributions of the observations and poorer fits. This is then reflected in,
  4. comparison of dates and their uncertainties calculated via two statistical methods using the same model, shown in the first column of panel (c). If the two approaches agree then we can have confidence in the robustness of the results. This is not well explained in the paper and I had to go back to the paper describing the diffusion-adsorption-decay model to understand what is shown The smooth blue curves are calculated using a Bayesian approach which evaluates the whole probability distribution, while the red lines use a maximum-likelihood approach and assume a normal distribution. The paper does not give ranges based on the Bayesian approach, but it looks to me that for CM-292 the two results are almost the same. For the other two bones the two approaches differ, both in the most likely date, which is younger in the Bayesian approach, and in the uncertainty range, which appears to be wider in the Bayesian approach. But all-in-all the values are fairly close, and  the differences in best-estimates are of no consequence given the uncertainties.
  5. From the data it is possible to calculate the ratio of 234U/238U that was supplied to the bone from groundwater before any radioactive decay took place in the bone. Comparing this to expected values can help test whether the model is correct. The authors state “Initial 234U/238U activity ratios calculated for bone subsamples span a narrow range (1.38-–1.50) that is consistent with modern shallow groundwater from the nearby Sweetwater River drainage (1.45–1.54)”. They do not provide any argument to show that these waters should be comparable to those the bones have been exposed to over a full glacial cycle, nor any indication of how 'nearby' this is. Nevertheless, if this is the right comparison, the closeness of the values is encouraging.
Overall the authors have carried out almost all the checks that could be done using uranium-series data, though they have not evaluated other elements in the bone which could have added to the robustness of the results. In terms of uranium-series dating of bone, this is one of the best-evidenced studies to date.

Luminescence dating



The supplementary material reports a series of luminescence dating studies by Steven Forman at University of Illinois at Chicago and Ronald J. Goble at the University of Nebraska. Neither of these contributors are authors of the article. These attempts were unsuccessful as the upper limit of dating at this site is about 70,000 years. It would have been nice to see more detail of why this upper limit is reached, including plots showing how close to saturation the dates are.

Stratigraphic context


Scientific dating methods don't exist in a vacuum. The dated samples come from a stratigraphic context, which can provide constraints and checks on the dates. The supplementary information describes the site as lying “within a 12 m-thick series of flat-lying, late Pleistocene fluvial sediments”, but the figures and descriptions of sediments only cover about 3.5m of that sequence. To check the dates it would be useful to have a figure of the wider stratigraphic context, showing stratigraphic correlations with nearby sites and previous estimates of the ages of the strata. Without this it is difficult to put what is effectively a single high-quality date into context.

The supplementary material does refer to a site report which may contain more detail. I've not been able to find this online but I did find a scan of the executive summary. This does not contain any further stratigraphic information but does state “Radiometric dating of ivory and soil carbonate from the quarry yielded dates of 335+/-35 Ka (thousands of years before present) and 196+/-15 Ka”. What are these dates? They do not feature in the paper or the supplementary information, but prima facie they differ greatly from the date proposed in the paper and it would seem that this radiometric data needs to be reported and shown to be wrong before we can be confident about the date of the site.

Conclusion


The Cerutti Mastodon site has one well-dated mastodon, but the lack of stratigraphic context and the hints of unpublished contradictory evidence weaken the robustness of the claimed date. I'm not so well placed to evaluate the evidence for human activity at the site, but the lack of formal stone tools, lack of butchery marks and selection of bones and teeth to smash do not look to be like other sites of this period. Further confirmation will be needed before this enters textbooks as the earliest human occupation in the Americas.