Radiocarbon Dates from Akrotiri: Problems and a Strategy

As illustrated in the paper entitled "Radiocarbon Dates from the Site of Akrotiri, Thera, 1967 - 1917" by H. N. Michael, 70 per cent of the ¹⁴C dates for the level of final destruction deviate or at least depart by more than one standard deviation from the expected period of destruction, 1550 - 1500 B.C. (Marinatos 1939; Marinatos 1971; Hood 1971).

Statistical expectations alone would suggest that only one-third of the dates should deviate by more than one sigma from the expected date.

Among the possible causes of these deviations are the volcanic nature of the site, sample collection, accuracy of ¹⁴C dating, and the reliability of MASCA correction factors, etc. The site is unusual in that there have been several volcanic eruptions and the surface is now covered with meters of pumice. We do not know how long the eruptions on Thera lasted nor whether or not gases, including older CO₂, emanated for long periods of time. We do know, however, that the mixing rate of the atmosphere on a world-wide basis is rapid, that is, of the order of two years even between hemispheres (Walton, et al. 1970; Tauber 1967; Lal & Rania 1966). Since Thera is now a windy island, one might assume that only vegetation directly in the path of emanating CO₂ would have been affected.

Sulerzhitzky (1970) reported radiocarbon measurements of near modem trees and charcoal at varying distances, even 4 km, from volcanoes on Kunashir and Simushir Islands, Russia. Apparent ages ranged up to 2280 years, even 2 km away.

This and other series did not help to establish possible irregularities in the local distribution of the carbon dioxide from volcanoes. As on Thera, one needs to know the distribution of volcanic gases and the duration of active gas emanation, both before and after the eruption.

In regard to the collection of samples, we assume that the excavators were well qualified to collect samples which represented what should be dated in the excavations. Many of the samples consisted of short-lived materials such as fava beans, grains of various kinds, and shrub branches. For the future, however, we would suggest that those taking the samples should be assisted by one of the radiocarbon scientists involved in their dating.

With regard to the accuracy of Thera dates from the MASCA ¹⁴C laboratory, we have taken a critical look at the standard oak calibration samples that were counted before and after and sometimes interspersed between the counting of Thera samples.

The counting of the oak samples was unusually consistent. This shows that there is no reason to suppose that any systematic or exceptional random error or errors have affected the dating of the Thera samples.

With respect to the general accuracy of ¹⁴C dating, one of the best examples from the University of Pennsylvania Laboratory is the study of the correlation of the Maya calendar with the Christian (Satterthwaite & Ralph 1960; Ralph 1961; and Ralph 1965). The problem was that the Maya long count system did not survive the classic period; and that part of the calendar which remained in use, i.e., the short count, repeated itself every 260 years. The site of Tikal was being excavated by the University Museum and the samples were collected from the remaining outer growth layers of vault and lintel beams which contained the calendrical hieroglyphs. We made a total of 33 counting runs for 10 different beams of Temple IV, and 17 runs for 6 different beams of Temple I. With the exception of one baam from Temple I, all counts were statistically consistent and supported an 11 - 16 (Goodman-Thompson-Martinez) Maya-Christian correlation. We assumed that the one deviant beam from Temple I had been reused. The statistical uncertainty in the counting for Temple IV was +/- 16 years and for Temple I, ± 21 years, but with the ± half-life uncertainty added, these figures became ± 34 and ± 37, respectively. 

Since the half-life of ¹⁴C has subsequently been changed from 5568 to 5730 years, we now know that the ± 30 uncertainty was not realistic. The original dates for Tikal were calculated with the 5568 half-life, but when revised to the 5730 half-life and corrected by the sequoia and bristlecone pine tree-ring dated series, the resultant dates are in good agreement with those calculated with the 5568 half-life.

This leads to the long series of over 700 samples, of dendrochronologically dated sequoias (Sequoia gigantea) and bristlecone pines (Pinus aristata) which have now also been dated by ¹⁴C. This series spans the period from A.D. 1850 to 5500 B.C. and may soon be extended to 6100 B.C. or earlier. The tree-ring dating has been done mostly by the Laboratory of Tree-Ring Research, Tucson, Arizona, and the ¹⁴C dating by the laboratories of the Universities of Arizona, California at La Jolla, and Pennsylvania. As seen in Fig. 1., there is good agreement among the three laboratories on the average but the scatter is greater than expected from statistical considerations.

Therefore, we have turned to various methods to reduce this scatter. In 1973 (Ralph, Michael & Han) we employed a third order polynomial to obtain the overall trend and nine-sample regression averaging to reveal the shorter term oscillations. More recently, we have turned to a sixth order polynomial (see Fig. 2) and in addition, we have used sinusoidal fits to study the shorter term oscillations. Some are shown in Fig. 3.

Again, because the mixing rate of the atmosphere is rapid, these plots and resultant tables of trees from California, U.S.A. serve as a means of correcting the dates from allover the world. Also, they provide a means of studying the possible correlations with the basic geophysical causes of these oscillations.

In this lengthy series of ¹⁴C dates for tree-ring-dated sequoias and bristlecone pines, as published in 1977 (Michael & Ralph) only 11 ¹⁴C dates among 226, or 4.9 per cent, were aberrant.

Another long series of ¹⁴C dates is that of the "King Arthur's" Round Table, Winchester, England, dated by the Harwell Laboratory. This great medieval table, 18 feet in diameter, has hung in Winchester Castle Hall since at least the mid-fourteenth century A.D. The date of its construction was unknown, but is now placed by ¹⁴C at A.D. 1234  ± 13. This date seems now to be confirmed by a tree-ring date of A.D. 1237 from a 23-board master curve related to the Winchester master dendrochronology curve. Among 124 separate ¹⁴C counts, only 6.5 per cent were rejected, either for possible random contamination, or because the counting results differed more than three sigma from the mean (only 2.4 per cent).

In other words, the dating of these samples of wood is considerably more consistent than that of the samples from Thera. It seems that the exceptional factor at Thera may have been the volcanic eruptions. It is also not inconceivable that some of the samples could have been mixed with earlier materials. Without a knowledge of the duration and precise locations of the eruptions and the consequent mixing of deposits, it is difficult to answer these questions.

In these circumstances, it is clear that the ¹⁴C dates presently available for Thera cannot be regarded as either final or satisfactory. We would like to suggest that a major programme of ¹⁴C dating should now be undertaken, designed to establish whether or not the ¹⁴C dating of Thera is consistent with the currently accepted archaeological dating for the occupation of Akrotiri and the eruption of "1550 - 1500 BC"

In particular we suggest that a series of perhaps 100 dates should be attempted, should be collected by the excavators working side by side with a member of the ¹⁴C laboratory or laboratories responsible for the dating. The samples should be taken from clearly defined, sealed, archaeological contexts and should seek to cover all the available materials: long-hard wood, shrubs, short-lived wood (eg. twigs), legumes, cereals, and bones (preferably burnt). If possible, an attempt should also be made to obtain samples both from the destruction deposits and from the earlier levels of the settlement.

It seems to us that the samples might each be divided for dating by two laboratories working in close collaboration and publishing together. We would be happy to provide the facilities of the MASCA laboratory for this programme.

-------------------------------------------

----------------------------------------