Hugely important new archæological technique not quite so important once actually published

I initially thought this might be one of the biggest stories in the study of the past for a long time. As you may already have seen from coverage at the BBC and on the webpages of the Royal Society and the Royal Society of Chemistry, all of which I found via variousworthyblogs, a team in Edinburgh and Manchester have developed a new technique for the precise dating of ceramics. To copy from from the Royal Society's page:

From the moment they are fired, ceramics begin to absorb moisture from the environment which causes them to gain mass. Using a technique they call 'rehydroxilation [sic] dating' researchers led by Dr Moira Wilson from the University of Manchester found that heating a sample of the relic to extreme temperatures causes this process to be reversed all the moisture it has gained since it was fired is lost again [again, sic].

The more weight a sample loses during heating, the more moisture there was to start with, and so the older the relic. After heating, Wilson and her team used an extremely accurate measuring device to monitor the sample as it began to recombine with moisture in the atmosphere. They then used a law to predict how long it would take for all the water lost in heating to be reabsorbed, and so reveal the true age of the sample.

Now when I read this, I immediately started digging for a text, because I wanted to know how close the dating was. That was before the Medieval Material Culture blog had pointed me at the BBC story, which suggests that the results are at least precise to within a year. That would be pretty heavy. At the moment archæological dating, at least for the Middle Ages, relies on various methods, all of which have their drawbacks. Here are the six I understand well enough to explain: if anyone can do thermoluminescence or radiometrics in comments, or contribute others I've missed, that would be much appreciated.

1. Carbon-14, relying on the fact that organic matter contains a small amount of unstable carbon which decays at a steady rate, meaning that you can measure the proportion of stable to unstable carbon, work out how much has gone and therefore how long it's been around. The problems with this are that you have to have suitable material at all, the data requires calibration to be realistic, and there are arguments about how to best do that, all kinds of issues over the selection of the samples, for the Middle Ages it is typically precise to 40 to 60 years, and that precision gets poorer as the sample gets older. It's also expensive, because of requiring a mass spectrometer, and takes a long time, which doesn't help.

2. Coins, which can very often be dated to within a year using historical methods, and usually at least to within a reign. The problems for this one are firstly, that your site must have turned up a coin, and secondly that numismatists can tell you the date of manufacture with alluring precision but what you really want is date of deposition, and in some cases that can be centuries later. In any case, sometimes the numismatists just don't know.

3. Correlation with historical sources, which is very unusual, subjective, despised by some archæologists as disciplinary miscegenation and failing to place the archæological data first, and of course makes you prisoner to all the problems that historians have to reckon with in their own line of work.

4. Stratigraphy, or relating objects to a relative chronology by the depth and layer (stratum) of a site in which they are found. Drawbacks: this only works on complex but undisturbed sites, i. e. if you only have one layer of occupation you won't get any relation between layers, but if you have them and someone's then buried a load of rubbish in them they won't be clearly distinguishable any more, because bits of them will have been shoveled into others and so forth. Also, of course, unless you have some other form of dating evidence too, it can only be relative and doesn't give you absolute dates.

5. Dendrochronology, or tree-ring dating, probably the best of the lot, in which the known fact that trees put on more wood in certain climate conditions than in others means that, by starting with existing trees and going back through older and older samples, a very complex histogram of 'thick' and 'thin' years has been built up against which pattern a known wood sample can be compared, much like a barcode. The only disadvantage to this is that you need surviving wood in relatively good condition, which tends to mean that it only works on waterlogged sites.

6. Lastly, typology, of ceramics especially but also other objects. This is the most common one, I think, for the historic period anyway, partly because it's oldest and also because it's least expensive. It relies on the idea that an artistic or 'craft' style (draw the line between those if you dare!) develops over time. So, experts have gathered corpora of material and tried to arrange it in a sequence of development into which other samples can then be fitted. This is of course immensely subjective and relies on a modern and Western judgement of what constitutes quality: words like `degenerate' and `barbarous' still turn up here when they've been banished from almost all other parts of the historical disciplines. They're allowed to survive because we need the dates they permit so much, but the whole thing is built on sand and whole sequences can be invalidated by subsequent finds without this making it back into scholarship of certain sites where such objects have been found. I myself, as you will have gathered, doubt that these dates can ever be 'valid' except in certain lucky and clearcut cases. I want the rôle of the individual as creator and agent to be allowed to make this all self-evidently impossible. Decisions to imitate, to distort, to exaggerate, to develop or to abandon styles and devices can't be reduced to a simple growth or waning of ability to do so. And again, of course, you need suitable objects from your site, but I think that's minor compared to the essential conceptual problem.

I speak with considerable pain here because people in my area and period of study specialised in a kind of coarse grey pottery which appears to have remained basically unchanged for about three or four centuries, they did not use and lose coins in such a way that they are easily found at the present time, and the region is almost never waterlogged—quite the reverse—leaving its archæologists with almost no useful dating evidence. So this new technique, if it not only makes typological dating redundant but provides a way of checking it, bringing precision and inarguable datapoints into a ceramic sequence, could really make a difference to me personally but obviously, also to thousands and thousands of other researchers.

So I wanted to read it. This took longer than it should have done. Despite the Royal Society's own press release, which is dated 20th May 2009 and says of the paper that it is"published online today in Proceedings of the Royal Society A", the paper actually only made it online on the 27th May. Before then I couldn't get it, and more to the point, neither could any of the people who blogged about it have done. So everyone was working from press releases and interviews and no-one actually checked to see if it were true. For shame, Internet! Well, never fear. Trading on my dubious reputation for trying to understand work with historical implications by scientists, I have got this paper and read it.¹ (And then, I admit, I asked for help from a real scientist with the difficult maths. Interdisciplinary conversation, you see, in the purest sense.) Having done that, I'm afraid we can't replace all other forms of archæological dating just yet.

The big thing that this research has achieved is the proof that fired-clay ceramics do soak up moisture at a steady rate over time. The title and the paper itself make it clear that this is what the team think is important about its work; they have shown that a process that many people would expect to be unreliable because of Brownian motion is in fact reliably measurable, predictable and rigorous. This is, moreover, irrespective of how much moisture there actually is around; that is, if it's underwater it soaks up no more than it would in the open air:

The rehdroxylation rate constant α is insensitive to water vapour pressure at all normal environmental levels. This is because the long-term uptake of water is both small and slow so that the ware demand is extremely modest; and because the rehdroxylation rate is controlled entirely by internal processes and does not increase when water is available in excess in the microenvironment.²

This being their aim, it is probably understandable if they have worried less about the historical problems, but that doesn't make them go away. The first is the dating precision. This is actually quite hard to get out of the paper. They don't show their working, and for all except one sample the data is only given in a caption to one of the graphs.³ That graph only shows the dates that they calculated for the samples on a logarithmic scale, moreover, so at the old end it's almost impossible to measure with any precision. An example calculation wouldn't have taken much space and would have made this paper far easier for someone outside their group to follow. Anyway, they had a Roman sample from a floor in London which had been archæologically dated to CE 50X160, and for that they got an age of 2001 years, and so a date of CE 8. (I have to extract the date from the BBC article, because it can't be measured in the paper; note however that the BBC article says that the archæological date was c. 2000 years old, which is not what the paper says.) Well, OK, archæologists have, as said above, got some space to be wrong in given available methods, but we have to recognise that the Romans weren't in Britain in CE 8 so an opus spicatum floor in London just can't date from then. London hadn't been founded yet! The result here must be out by at least fifty years, whether it fits into an appropriate error bar or not. (I'd like the BBC to have spotted this really, but never mind.)

It seems then that although the method inherently gives a value and not a range, there is some variation between results. Their only discussed example case is a piece of clay brick from the King Charles II Building in Greenwich, UK, which was built (historically) in 1664-69 and rebuilt in the 1690s, from which they processed five samples. From that they got a date of 1691 ±22 years, and therefore decided that their sample must be from the reconstruction. The error margin is to one standard deviation, but without having their data it's impossible to back-form the actual results from that; it could be that one or two of the five results were way way out and could easily lap into the original building. Five more of which three fluked early could make that an apparent certainty. So I don't think this experiment actually provided the historical input that the authors think it did, though it shows the potential.⁴

Worst of all, this process has in it one artificial value, because the rate of moisture uptake varies with environmental temperature. The authors therefore use a constant for mean lifetime temperature that they got from another paper,"for mean annual ground and air temperatures at several stations in southern England". But the paper they cite has a temperature series going back only to 1772.⁵ Now, you may have heard mention of a thing called 'the medieval warm period'. This is a historical amelioration of temperature in Europe between, roughly, the tenth and twelfth centuries. This probably decreased rainfall and other sorts of weather bad for crops, therefore boosted agricultural yield, pumped more surplus into the economy, fuelled demographic growth and arguably deliquesced most European societies to the point where they changed in considerable degree. However, because of the current debate on climate change, it has become a ball to kick around for climate 'scientists', those who wish to argue that we're not changing the climate pointing to it and ice coverage in Norse-period Greenland (which was less than there is currently despite less carbon dioxide in the atmosphere then), while those who wish to argue that we are changing the climate (and, almost always, that this relates to CO₂ output, which does seem like a weak link in the argument) dismiss it as legend or scorn the very few and unscientific datapoints, not really caring that the historical development of European society in the ninth to eleventh centuries just doesn't make sense without this system change from the ground. None of these people are medievalists and they're not trying to prove anything about the Middle Ages, so it gets messy, but there is a case about this temperature change that has to be dealt with. This obviously has an impact on this research. If the sample were old enough, the errors and change probably ought to balance out. But if it were, from, say, the eighth century, then the moisture uptake in the four or five subsequent centuries would be higher than expected from the constant that this research used and the figure would be out, by, well, how much? The team didn't know:

The choice of mean lifetime temperature provides the main other source of uncertainty, but we are unable to quantify the uncertainty in this temperature at present.⁶

We, however, need to know how far that could knock out the figures. Twenty years? More? It begins to push the potential error from a single sample to something closer to a century than a year. That is, the margin of historical error (as opposed to mathematical error) on this method could be worse than that of carbon-dating, and we don't actually know what it is.

On the other hand! to do this method you need: a furnace capable of maintaining 500° C for four hours, a microbalance capable of weighing grams to four decimal points, about 3 g of sample material and someone who can do difficult maths. I am assured that this is a lot cheaper and simpler than carbon-dating and takes much less time. Since potsherds are often larger than 3 g, also, you could easily do five or ten samples at once and thus reduce at least one source of error, though the temperature assumption will continue to poison the figures. All the same, it would give you figures that you could trust at least as much as anything else short of dendrochronology and since it works on a different, and more common, sort of evidence that's a very useful check to have. In some areas, too, this will allow a dating of some kind where one simply wasn't possible before. And, as the authors argue, it may also be possible to reverse the method. That is, samples of known age can be used to pin down the dangerous variable, that of mean temperature, and:"such an estimate may itself have value, for example in climate studies".⁷ Well, quite, and my helpful scientist pointed out that with enough such samples you could start building a library of known temperatures, like the dendrochronology master chart, that would make every subsequent test done the more rigorous. So after a few decades of dating ceramics like this it really might reach genuine precision. Aceramic areas will still be blanks, of course, and as yet it's not going to replace any of the previous methods, except maybe typology. If it does that, it will threaten rather a lot of jobs, but it will also bring us a lot closer to historical understanding. What's the price of that, then?

1. Moira A. Wilson, Margaret A. Carter, Christopher Hall, William D. Hoff, Ceren Ince, Shaun D. Savage, Bernard McKay & Ian M. Betts,"Dating fired-clay ceramics using long-term power law rehydroxilation kinetics" in Proceedings of the Royal Society A: Mathematical, Physical & Engineering Sciences, published online 27 May 2009 in advance of the print journal, doi: 10.1098/rspa.2009.0117, temporarily paginated.

2.Ibid., pp. 3-4.

3.Ibid., p. 6 fig. 4.

4.Ibid., pp. 4-7.

5.Ibid., p. 5, citing D. E. Parker, T. P. Legg & C. K. Folland,"A new daily Central England temperature series, 1772-1991" in International Journal of Climatology Vol. 12 (London 1992), pp. 317-342, doi 10:1002/joc.3370120402.

6. Wilson et al.,"Dating fired-clay ceramics", p. 7.

7.Ibid.