Colavito concludes on his blog
Quote
Regardless of the authors’ correctness on the source of the meteoric fragments, their conclusion cannot be correct because the Hopewell did not enter a terminal decline after their proposed impact date of c. 255-300 CE but flourished for another 200 years.
First thing of note is that Colavito has truncated the calibrated radiocarbon dates for the apparent airburst evidence by around 83 years. Tankersley et al. give the age range of 252-383 cal AD (calibrated AD). Even so, there is still a good difference between this age range and the decline of the Hopewell culture circa AD 500. However, I feel that there are a number of issues with the radiocarbon analyses of Tankersley et al.
The authors state that they had 29 radiocarbon dates from four sites which they used Bayesian analysis to obtain date of the “Hopewell airburst” event. They do not though provide any details of the dating model that they used to obtain their final, although they do provide the radiocarbon ages of the samples used in the supplementary material. The lack of information makes it difficult to assess the robustness of their age model, however, from what I can see (and it is quite possible I am missing something) they did not do anything more than take a weighted mean of the radiocarbon ages of samples taken from the nominative “airburst-layer”, and calibrating that mean. By doing this they are actually only using 20 of the 29 radiocarbon dates from the sites, and in the absence of the radiocarbon ages of the strata above and below the airburst layer.
Now, taking the weighted mean is not always the wrong thing to do, BUT typically you would only use a weighted mean on samples that you know have, or strongly suspect have shared the same radiocarbon reservoir. For example, you would use a weighted mean on multiple samples taken from the same bone, or from the same portion of a tree. Or if a sample was sent to different radiocarbon labs. Or if you had an assembly of short lived samples from a secure context (like a sealed jar of seeds). What you should not do, unless you have an extremely robust justification for doing so, is take the weighted mean from different types of samples, or from samples that may or may not be associated with each other.
That unfortunately is what Tankersley et al. appear to have done in this case. Take for example the artefacts dated from the “airburst-layer” from the Turner Earthworks. We have in this ensemble 4 wood charcoal dates, 1 bone collagen date, and 2 charred grape seeds. And this is a problem. The grape seeds would be short lived samples, which is great, but we have no context where in the stratum they were found, or if they should be associated with each other (i.e. they both grew in the same year). We have bone, which has undergone carbon exchange not only in life, but also continues to do so for a decade or more (depending upon species it belonged to) after the creature’s death. And then we have the wood charcoal. This is a big issue, since you could well have the old wood effect. The only part of a tree that exhchanges carbon with the biosphere in a growing season is the outer ring. For all intents and purposes, the inner rings are “dead”, and so their radiocarbon signature is that of the past. What this means is, if you were to radiocarbon date a trunk or branch of a tree, you would find that the inner rings would be older than the outer rings (which is not that astonishing really). When you find charcoal, most times you do not know from what part of the tree it came from. Is it wood from the outer rings, or wood from core, or rings in-between? Invaribly when you find charcoal from the same archaeological layer, even if you can link the charcoal to the same time it was used, the dates will not necessarily tell you the time it was used, only when that wood last grew. Invariably this will give a large temporal range, and this range will likely be much older than the age it was used as fuel. It would be tempting to take the weighted mean of an ensemble of charcoal from a site, but again, the samples will likely have shared different carbon reservoirs at different times, even if they are all from the same tree.
But what all the above means is that realistically, you cannot with any confidence take the weighted mean of the collection of charcoal, bone, and seeds, and claim it to be robust. The true calendrical age of the charcoal for example could be 200+ years older than the bone, which itself could be decades older than the seeds.
It seems to me that Tankersley et al. have taken the weighted mean of samples from the “airburst layer” from four geographically separated sites (one of which only has one radiocarbon date)”. This ensemble of samples consists of in total; 13 carbonised wood samples, 4 bone samples, and 3 short lived samples (2 carbonized grape seeds, and nut shell), resulting in an age range of 252-383 cal AD. This is simply not a robust method. Even if this was robust, then the calibrated dates should be seen as being the EARLIEST possible dates, since there will inevitably be an old wood effect. In this respect, the age range could be even younger, and maybe approach AD 500, coincident with the Hopewell decline.
There are of course better models that one could employ using OxCal. One such model is the so called Tau_Boundary model. This model takes an ensemble of radiocarbon ages and assumes that they are exponentially distributed in time, with more dates falling closer to the younger boundary than the older boundary. If I use such a model for the Turner Earthworks and the Jennison Guard Village sites, which Tankersley et al. state are the better sites because
Quote
Radiocarbon ages from the Turner (N = 7) and Jennison-Guard (N = 8) sites are considered high-quality because they were obtained directly from the burned carbon-rich habitation strata, they have the highest number of ages per stratum, the smallest degrees of uncertainties, and they are associated with a plethora of temporally distinctive artifacts.
I get calibrated dates of Turner Earthworks = 253-445 cal AD (95%) and Jennison Guard = 360-535 cal AD (95%). Again one would need to take into consideration the old wood effect. The point is that these dates could easily extend into the late 5th/early-6th century, and could be contemporary with the Hopewell decline. Applying the same Tau_boundary model to the Marietta site though gives a calibrated date of 252-748 cal AD. The Miami fort only has one radiocarbon date and this gives a calibrated range of 116-641 cal AD. Again we must consider the old wood effect.
That’s all quite a spread of calibrations from samples from a strata that are all meant to be contemporaneous to each other. Without knowing much about the Hopewell, or the typology of their pottery and artifacts, and if what if any sherds are mixed in with the “airburst strata”, one could make the argument that perhaps the airburst strata post dates the Hopewell decline.
Other models will give differing results, and that is the important point (regardless as to whether my Tau_bondary model is applicable). If Tankersley et al. have used a weighted mean, I don’t think we can draw any firm conclusion on the date of the supposed airburst, as their radiocarbon analysis is not robust enough to firmly claim it occurred sometime in the 3rd-4th century.
Finally, there is another point I take issue with in Tankersley et al.’s paper. They state that during the interval that their radiocarbon dates cover, there were 69 near earth comets recorded. The implication to be read into this is that perhaps one of these comets caused the airburst. The fact is that of these comets, only a handful of the observations have enough observations for astronomers to compute their orbits (and a number of these are Comet Halley, and the visit in 374 was relatively close at around 14 million kilometres (give or take)). The fact is all we can say is that during the radiocarbon interval, 69 comets were observed, but we have no clue as to whether any of them came close to earth, or even crossed our orbital path to leave debris for us to collide with.
That being said, the node of Comet Halley (the point where it crosses the ecliptic plane) was close to the Earth’s orbit leading up to the 5th and 6th century, even though the comet did not come near to Earth at these times, so it is possible that debris from it could have caused an impact, but we need much more evidence to support such a speculation.
Jonny
The path to good scholarship is paved with imagined patterns. - David M Raup
