Radiocarbon dating is a method of determining the age of organic material (up to about 50,000 years old) by measuring the decay of radioactive carbon-14. It was developed in 1948–9 by Willard Libby at the University of Chicago.
Some elements have more than one type of atom, called isotopes. Carbon has three main isotopes, carbon-12, carbon-13 and radioactive carbon-14. The unstable carbon-14, created in the Earth’s upper atmosphere, enters the food chain along with the stable carbon isotopes. carbon-12 accounts for something like 99% of the total, and carbon-14 a minute fraction. The carbon-14 is in a constant state of decay but, as long as an organism is alive, ingesting more carbon, the balance between carbon-12 and carbon-14 remains stable. When the organism dies, however, new carbon is not being taken in, and so, as the carbon-14 decays, the ratio of carbon-12 to carbon-14 changes. The half-life of carbon-14 is 5,730 years. This means that, after 5,730 years, half of the carbon-14 will have gone. Therefore, the time that has elapsed since an organism’s death can be calculated from the proportion of carbon-14 left in a sample taken from its remains – though in samples older than about 50,000 years, there will be insufficient carbon-14 left to provide reliable results, and, conversely, recent samples will show too little decay to provide reliable results.
But, it turns out that the proportion of carbon-14 has varied significantly during the past 50,000 years, and so calibration curves have had to be developed (and, indeed, are still being developed) to compensate.[*] Consequently, radiocarbon dating does not provide pinpoint accuracy – the calibrated date is actually a probable date-range. Since the mid-1990s, however, ‘Bayesian statistics’ has been increasingly employed to combine the radiocarbon dates of samples from an archaeological site with other pertinent information – for instance, stratigraphy (i.e. the position of samples in the archaeological layers), or typology of artefacts – to greatly reduce the estimated date-range of the activity represented by those samples. A few centuries might be narrowed to a few decades.