Think Before You Radiocarbon

Recently I organised a few days' excavation that didn't turn up the kind of stuff I was hoping for. Still, I brought some materials home that may serve to shed some light on what exactly it was we dug into. All those nondescript little pits, all those sooty hearths full of cracked stone -- when were they made and used?

Enter radiocarbon. This dating method works on anything organic, that is, anything with carbon in it. Running one sample costs about $500, so you have multiple reasons to be smart about which samples you send to the lab. I thought my thinking about this might interest you, Dear Reader.

We dug 175 sunken features, but I don't have 175 samples. Most features yielded no datable material (in these situations, usually charcoal, other charred plant remains or bone). Of those that did, many also contained modern junk identifying them as recent refuse pits. I don't want to spend any money dating them.

I have about 20 samples, and I've decided to date five. That's not too bad for a 1000 square meter trench that contained nothing that appears relevant to my research project. I'm of course only running samples of which I know exactly what stratigraphic context they came from. I'm avoiding a few iffy ones that may have been contaminated, such as one that was collected from the sieve instead of from the feature's section. I'm trying to cover different kinds of feature (hearth, posthole, pit), though charcoal is of course most abundant in hearths. And I want to spread the dated features out across the trench in order to catch any horizontal variation.

Radiocarbon dates the moment when the tissue concerned stopped receiving carbon from the environment. For a leaf, this is the moment it stopped photosynthesising. For the soft tissues of animals (who, through the food chain, receive carbon taken from the air by plants), it's a matter of months. For bones, a bit more than that. And for dentine, the interior of teeth, it's years or decades.

Charcoal is tricky. Every year ring in a tree has a different radiocarbon date. Chop down an old oak, sample the centre of its trunk and some of its leaves, date both samples, and you'll get a discrepancy of centuries. This is because once a year-ring has formed, it ceases to receive carbon from the living parts of the tree. Therefore, I've left my selected charcoal samples to a wood anatomist. Often he can select young twigs or pieces of bark, with a low intrinsic age. Second best, he can select bits of charcoal from a tree species with a short maximum lifespan. Oaks live for centuries, but alders and aspens mostly don't, so they're better. Cereal grains and other seeds are excellent, no intrinsic age at all.

On the samples go to the radiocarbon lab, where they're cleaned, processed and transformed into a graphite coating on little brass plugs. Then some of the graphite is oxidised to CO2 and sent into a particle accelerator. Heavy unstable carbon-14 lands on one detector, lighter and stable carbon-12 on another, and the ratio between them shows how long time has passed since that tissue stopped receiving carbon.

Or it would, if the availability of carbon-14 in the atmosphere were constant. Which it is not. But that can be compensated for.

Back to the tree rings. In warm years, all the trees lay down thick rings, and conversely in cold years. Therefore, I can convert the ring thicknesses of a tree I just cut down into a series of figures, like "thick - thick - medium - thin - thick - medium - thick" etc. With large enough a piece of wood (>50 year rings), I get a unique sequence of thickness estimates that has never occurred before or later in tree history. Armed with my graph (or, actually these days, my database and stat software), I then seek out an old building and saw off a chunk of wood that provides me with a similar dataset. When I find the part where the two datasets overlap ("thick - thick - medium - thin - thick" etc.), I end up with a combined dataset that reaches way back before the original tree I cut down had started to grow. And this process kan be, and has been, repeated until we have tree-ring datasets that extend thousands of years into the past.

We know exactly what year each of those thousands of tree rings formed. This allows us to cross-check their radiocarbon dates, which provides us with a calibration curve. This curve allows us to translate from apparent radiocarbon dates to actual calendar dates. Anybody can to this at home, the best Oxonian software and most recent datasets are freeware.

Without calibration, you miss your actual date by a thousand years in many parts of the past. Around the introduction of agriculture to Scandyland, c. 4000 BC, for instance, the error is a full millennium.

The need for calibration also means that the precision of radiocarbon is not uniform across the millennia. In the 7th century AD, the calibration curve has a shape that allows highly precise dating, e.g. AD 625+-15 years. In other eras, e.g. about 4000 BC sad to tell, the calibration curve forms plateaux where a range of different radiocarbon figures map onto the same calendar date after calibration, e.g. 3950 BC. This means that the entire Scandy process of neolithisation gets compressed, looking like it happened in 3950 BC. There simply aren't any possible radiocarbon values that will point, after calibration, at 4050 BC or 3850 BC.

Now, what about those five samples of mine? I hope to get them dated in December once some funding paperwork has gone through. I expect them to land in the interval 500 BC to AD 400, because that's the usual date of this kind of site with loads of pointless pits and hearths. However, I wouldn't be surprised if I got one date at about AD 1900 and another one in the 6th Millennium BC, because those are the periods from which the dig actually produced any artefacts. I'll get back to you on the subject when I know.

Questions?

Update 17 October: More fresh radiocarbon goodness in Swedish from Ãsa of Ting & Tankar.

[More blog entries about archaeology, radiocarbon; arkeologi, kol-14.]