Neutron Activation Analysis of Pottery from Thera

The concentration of 14 elements were measured and, with the aid of a clustering technique, groupings were obtained in broad agreement with those of Williams. Differences in detail occurred, however, which pointed to the useful complementary function of the two methods of analyses.

INTRODUCTION

In his paper on petrological examination of Theran pottery, Mr. Williams referred to the desirability of collaborative studies in assessing the provenance of geologically-based artefacts. Of particular interest is the comparison between identification using the basic chemical elements and the nature of inclusions present in pottery fabric. The former analyses may be regarded as "total" in that unless inclusions are specifically excluded, the elements present in them will contribute to measured concentration levels and no account is therefore taken of whether the inclusions are incipient in the original clay or are deliberate additives. The petrological approach may, with the aid of an appropriate geological study, provide enlightenment. It seems reasonable to assume, however, that, since we are searching for the characteristics of a manufactured product, the total assemblage of components should be considered, provided our sample for analysis has not been distorted by the unrepresentative presence of, say, large inclusions. Even the presence of components which may be relatively inert to the methods of analysis used, such as some calcareous grits to neutron activation analysis, will merely serve to "dilute" the apparent concentrations of elements measured and thus provide some characterisation.

The papers contributed by Einfalt, Jones and Noll have all pointed to the desirability of analysis by differing techniques. Experience in the investigation of English and Medieval ceramics (Aspinall 1977; Hawkin 1977) has shown that multi-element analysis, combined with appropriate data treatment, can be a powerful tool for artefact discrimination. Such methods have, indeed been widely employed in ceramic studies (see for example, Bieber et al. 1976; Davidson & Mc Kerrell 1976). The opportunity was taken, therefore, to subject the pottery sherds, supplied to Mr. Williams by Dr. Doumas, to neutron activation analysis and treat the results by routine cluster analysis method.

METHOD

Thirty-eight sherds of the original collection were examined. A small area of "surface" was removed from each specimen and the underlying fabric sampled using a diamond-headed drill to provide approximately 100 mg. of pottery powder. No particular precautions were taken to exclude small (relative to the sample volume) inclusions from the drilling. Each weighed sample was packed in a small polythene container and the total batch, together with a pottery "standard", was irradiated in the Herald reactor, A.W.R.E. Aldermaston, for 24 hours at a thermal flux density of 1.8 X 10¹² n cm^-2 sec^-1.

Subsequent examination of the γ-ray spectra was carried out at Bradford using the conventional high resolution techniques of activation analysis. Nineteen elements were recognised as present in all or some of the specimens; however only fifteen of these were regarded as having sufficient reliability to merit quantitative treatment.

One element, scandium, was chosen as a 'normalising' factor and, in data analysis, all other element concentrations in a particular specimen were referred to the scandium concentrations. Thus fourteen 'normalised' concentrations were used in data treatment.

RESULTS

All element concentrations were given the same weighting and subjected to the cluster analysis treatment of Ward (Sneath & Sokal 1973). Earlier work (Hawkin 1977) had suggested that English medieval ceramics from geographically close sources could be distinguished better by this approach than by other accepted cluster methods. No assumptions of "local" or "imported" origin were made about the samples so that the dendrogram (fig. 1) appearing from the analysis should be regarded somewhat qualitatively in terms of statistical reliability. A true statistical treatment must await larger sample numbers from assumed sources.

DISCUSSION

The dendrogram is striking in that two clearly defined major groups emerge and it is tempting to conclude that the "local" and "imported" forms have been resolved. A closer study of the dendrogram, however, reveals a fine structure to the two main groups which it is interesting to compare with the conclusions reached by Williams.

In general the two main groups are distinguished from one another as follows.

The first major group (fig. 1), which is predominantly "imported" according to Doumas' description, is characterised by low rare earth, and alkali metal concentrations and high cobalt and chromium concentrations relative to the second group. The observations on cobalt and chromium are in good agreement with those of Jones and Noll, given in earlier papers, as characterising imported wares.

A study of the structure within each main group shows that these broad generalisations are not always valid. Within the "imported" group (fig. 2), it can be seen that samples appear which were regarded by Williams as "local". Indeed three of his Group 1 specimes (8, 17, 6) form a distinct set. It is found that, in fact, there is a common feature of low cobalt and rare earth concentration amongst them whilst specimen 8 also shows a low chromium concentration. Clearly it is feasible that these specimens may be of local origin but from a source other than that of Williams' Group 1. Specimen 2 on the other hand exhibits all the characteristics of an "imported" ware. The analysis has not confirmed the separate Groups 7, 9, 10 and 11 of Williams and merely classes them together.

The second predominantly "local" group (fig. 3) is not so tightly defined as the "imports" group and can be readily subdivided into smaller sets. Seven of Williams' Group 1 together with his Groups 2 and 5 come together with the single member of Group 8 (sample 28). The limestone inclusions observed in the Group 5 samples by Williams do not, apparently, influence the basic element pattern, whilst his interpreration of sample 28 as a local copy of Group 6 appears to be confirmed. Three members of Williams' Group 6 cluster well before joining the main "local" group, but the fourth, specimen 23, is firmly established amongst the "imports", with high cobalt and chromium content. It thus appears that a distinctive source of unknown origin may be ascribed to the three samples 21, 22 and 27. Williams' Groups 3 and 4 come together in the analysis and are characterised by high sodium content. They cluster loosely with specimens 13, 20 and 3 from his Group 1, and all show significantly high europium content relative to other rare earths. Such anomalies in europium are usually regarded as having a geological significance, implying a distinct source for these specimens.

CONCLUSIONS

Multi-element analysis of the pottery collection from Thera suggests that we are concerned with five sources of origin. Main groups of local and Cretan products, which agree broadly with the findings of Williams, are defined, but three other characteristic groups also appear. The origins of these cannot be ascribed until further source specimens are analysed. It would be appropriate at this stage to undertake analyses of Cycladic and mainland pottery and to investigate further the clay-rich areas of Thera.

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

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