A Study of Middle and Late Cycladic Pottery from Akrotiri

The pottery classes analysed from Akrotiri consisted of ceramics commonly abundant at the site and divided either according to the shape (e.g. stirrup jars, cooking pots, nippled ewers) or according to the type of ware (e.g. Middle Cycladic bichrome, Cycladic white, burnished ware).

The statistical analysis results showed that the majority of the Akrotiri pottery formed two compositional groups, which represent two different clay sources. The Phylakopi group was chemically related to those of Akrotiri due to the similar geology of the islands, but they were clearly separable by multivariate analysis. The Phylakopi group contained a small number of pots found at Akrotiri. There was also importation to Akrotiri and Phylakopi from Minoan Crete, which dominated the Aegean during the Late Cycladic period, though this never exceeded local production. Finally, about 20% of the Akrotiri specimens analysed did not cluster with any known provenance group, but probably came from various sources as a result of the many contacts Akrotiri had developed.

In addition some characteristic samples were examined, using the scanning electron microscope, in order to obtain information concerning the technology of manufacture.

INTRODUCTION

The excavations at Akrotiri have brought to light large quantities of pottery. When the inhabitants evacuated the town, they took only valuable possessions with them and abandoned this very important testimony to their life. Thera played an important role in the Aegean during the Middle and Late Bronze Age and its pottery is of great value in helping us to understand the Cycladic civilization.

The majority of excavated pottery belongs to the Middle and Late Cycladic periods. The paste texture is semi-frne and it is either matt-painted (black or brown) or burnished. There are of course exceptions which usually characterize some wares - i.e. cooking pots which are not classified archaeologically in the main body of Akrotiri pottery.

The study of pottery has provoked many questions concerning relationships among the clays used on various wares within Akrotiri and also within the other Cycladic islands, Crete and the mainland. Other questions have arisen concerning the technology of manufacture and the origin of the raw material.

In this study, attempts were made to answer some specific questions dealing with the chemical characterization using neutron activation analysis. In general, chemical analysis has been proved to be a useful tool in the investigation of problems dealing with pottery relationships. The interest in similar studies about Thera goes as far back as 1869 when M. Fouqué published an article in the Spectator concluding that the pottery should have come from abroad, quoted in Harbottle's article (1976). All the recent analytical work on Cycladic and Cretan pottery has been reviewed by Jones (1986) and also reported at the 'Thera and the Aegean World' Congress in 1978. The conclusion is that local pottery is distinguishable from the Cretan and mainland imports (Jones 1978; Aspinall 1978; Einfalt 1978). However it was noted that there was no clear distinction between Theran and Milian pottery (Jones 1978).

The problems we dealt with and the questions that were posed can be interpreted as follows:

1. The breasted ewer is a typical Cycladic form, very common in Akrotiri and Phylakopi. It has been suggested that Phylakopi was the production centre (Scholes 1978) and this hypothesis was tested in order to decide whether the breasted ewers found in Akrotiri were imported from Phylakopi or made locally.
2. Stirrup jars were mainly used for the transport of fluid commodities (wine, olive oil) and therefore were the most likely to travel. The provenance of the stirrup jars found in Akrotiri would automatically give evidence for the contacts of the site. One of the possible areas of import was Knossos (Doumas 1983).
3. Cooking pots, unlike the other Akrotiri pottery, are characterized by visible amounts of mica. Naxos is considered to be a production centre for this kind of pottery and a very large number has been found there. In addition, Naxian clay is rich in mica, and for this reason it was examined to discover whether the cooking pots from Akrotiri were imported from Naxos (Doumas 1983).
4. Middle Cycladic pottery is quite abundant in Akrotiri (Papagiannopoulou 1987). Most of the motifs and forms were considered to be the prototypes for the later developments (breasted ewers).

Questions, like whether the various Middle Cycladic wares were chemically related, whether there were any exchanges between Akrotiri and Phylakopi and whether the same clay was used in both Middle and Late Cycladic periods in Akrotiri, were investigated. Also, some aspects of the technology of manufacture of the Middle Cycladic wares are discussed, based on results obtained by examination of the pottery with a scanning electron microscope.

The pottery classes that were analysed from Akrotiri consisted of those commonly found at the excavation site, namely:

1. Breasted ewers (29 samples). These are classified in the major group of 'light fabric'. The paste texture is semi-fine.
2. Stirrup jars (10 samples). Their general appearance and their clay texture vary sufficiently to suggest different production centres.
3. Tripod cooking pots (21 samples). Their texture is semi-fine to coarse.
4. A selection of local and imported Late Cycladic sherds from Akrotiri (25 samples). These sherds were used to form a control group and compare it with the various pottery classes.
5. Middle Cycladic sherds (97 samples) of rather semi-fine texture which were stylistically divided into four groups: a) cycladic white (31 samples) with matt-black decoration on a buff background, b) bichrome ware (27 samples) with black or red decoration on a pale buff background, c) red burnished (20 samples) and d) dark burnished (19 samples) both without decoration, but with a thin slip applied on the body and then burnished.

Some local Cycladic white sherds from Phylakopi (12), stirrup jars from Knossos (15) and cooking pots from Naxos (22), were also analysed.

CHEMICAL ANALYSIS

Instrumental neutron activation analysis was carried out for the determination of twenty minor and trace elements in all the samples taken from the sherds and the pots. Samples which were to be activated were obtained by the use of tungsten carbide drill-bit. Two standards were employed: an International Atomic Energy Agency Certified Reference Material, SOIL 5, as well as a U.S. Geological Survey Standard Rock, GSP 1. Samples and standards weighing about 100 mg were sealed in quartz tubes and irradiated for three hours in the Demokritos swimming pool reactor using the rotation system where the neutron flux density is about 2 X 10¹³ n. cm^-2 sec^-1. After irradiation, gamma spectra of standards and samples were measured twice. The first count followed a cooling period of six days for radio nuclides with half-lives ranging from 0.6 days to 7 days in order to get results for As, La, Lu, Na, Sb, Sm, U and Yb. A second count was carried out after a cooling period of approximately three weeks for the long-lived radionuclides to get data for Ce, Co, Cr, Cs, Eu, Hf, Fe, Rb, Sc, Ta, Tb, and Th (Kilikoglou 1988).

RESULTS AND DISCUSSION

The first step of the data-set analysis was the subjection to cluster analysis of a matrix of Euclidean distances that were used as a measure of inter-sample similarity. This first step included all the Late Cycladic samples from Akrotiri, the cooking pots from Naxos, the stirrup jars from Knossos and the Late Cycladic control group from Phylakopi. The result of clustering was the separation of samples into several groups, the significance of which was further tested. Where there are strong inter-elemental correlations in the data, numerical analysis employing Euclidean distances does not usually describe successfully the existing relationships (Rohlf 1967). In the present case the correlations among many elements are quite high. For example, among the rare earths the coefficients are 0.75-0.90 and between Fe and Sc, 0.89. For this reason the cluster analysis results were considered only as an indication of grouping within the data. The final assignment of the specimens to the groups was made on the basis of calculations of multivariate probabilities that each individual specimen might belong to each of the groups that were formed by clustering. All the calculations were done with the program ADCORR (Sayre 1977).

The general impression that was given by the study of the cluster analysis dendrogram was that each group contained specimens predominantly found at individual sites. Only the samples from Akrotiri and Phylakopi did not separate but they all appeared in the same cluster. As will be shown, further statistical treatment proved that they could be separated. Refinement of the cluster analysis groups was made after the calculation of the multivariate probabilities. The mean elemental compositions for the defined groups appear in Table 1.

The majority of the Late Cycladic selection of sherds from Akrotiri belong to the same compositional unit. Fourteen of the twenty-one analysed formed a single statistical group which shared the purpose of a control group. From the rest of the local sherds (seven), three grouped with the samples from Knossos, one with Phylakopi and three remained unclustered. Twenty out of the twenty-nine analysed breasted ewers matched with the Akrotiri control, which suggests local provenance. As a result, a large group was formed containing the samples from the control, the twenty breasted ewers and a single sherd from Phylakopi. This group (Group 1), represents the main body of the Akrotiri Late Cycladic samples that were analysed, except the cooking pots and the stirrup jars which will be discussed separately.

The samples from Phylakopi were separated from Group 1 and nine of them (out of twelve) formed a single chemical group (Group 2) along with four breasted ewers found at Akrotiri. The clear separation of the four breasted ewers from the rest which are included in Group 1 and their chemical match with Phylakopi local sherds, indicates that they were imported from Phylakopi and confirms the Late Cycladic contacts between the two islands (Davis 1979). The distribution of the Akrotiri breasted ewers in both Group 1 and Group 2 suggests that they were mainly made where they were found and some of them (four out of twenty-nine) were imported from Phylakopi. Their production might have been influenced by Phylakopi's prototypes, giving the impression that Phylakopi was itself the only production centre (Scholes 1978).

The discrimination between Group 1 and Group 2 on the basis of their chemical compositions is not immediately obvious when the raw elemental concentrations alone are compared. As has been noted elsewhere (Jones 1978), the chemical patterns of the pottery from the two islands are very similar. The confusion arising from consideration of just the raw data was demonstrated in the dendrogram of Fig. 1 which contains all the specimens of Group 1 and Group 2. Samples with codes beginning with 'A' come from Akrotiri and belong to Group 1, except ABE038, ABE111, ABE114, ABE115 and ALS066 which belong to Group 2. All samples with codes beginning with 'F' come from Phylakopi and belong to Group 2. In this dendrogram there is a small tight cluster that contains only Akrotiri pottery and a larger one that contains everything else. The points of relative similarity at which the small subclusters are joined together does not allow the discernment of clear distinctions. The same set of specimens was then subjected to cluster analysis using not the raw concentrations as previously employed, but the standardized coordinates for each data point with respect to the core group, Group 2. Fig. 2 is the dendrogram of this clustering. It can be seen that two clearly separated clusters were formed. The first contained only the samples of Group 2 and the second only the samples of Group 1. The standardized coordinates were the projections of all the data points upon the characteristic vectors composed of linear combinations of the concentrations of the elements Sm, Yb, Lu, Ce, Hf, Cs, Eu, and Cr, in turn derived from the inverse of the variance-covariance matrix of Group 2. The most discriminating elements in this comparison proved to be Sm, Lu, Yb, and Cr. When bivariate plots were prepared for all the possible pairs of elements using the Group 1-Group 2 data set, however, there was a considerable overlap among groups. Only the multivariate coordinates had the necessary resolution to distinguish Akrotiri from Phylakopi. This is a typical situation where similar groups may be separable in multi-dimensional space.

The stirrup jars from Knossos were analysed in order to compare them with similar ones found in Akrotiri. Eleven out of the fifteen Knossos stirrup jars along with two out of ten Akrotiri stirrup jars and four Akrotiri imported sherds formed the third compositional unit that appeared within the analysed specimens (Group 3). The contribution of the Akrotiri sherds to Group 3 obviously indicates its contact with Minoan Crete. Also two of the ten stirrup jars from Akrotiri matched up to Group 3, and it can be suggested that Knossos was one of the cities that was supplying Akrotiri with goods, as stirrup jars were used for the transport of fluid commodities. The rest of the Akrotiri stirrup jars did not cluster with any of the groups.

The cooking pots from Akrotiri formed a separate chemical group (Group 4), containing fifteen of the twenty analysed. The other five remained unclustered. These vessels form stylistically a unique group of pottery because of the visible amounts of mica present in their clay, while the rest of the Akrotiri pottery is characterized by the absence of visible amounts of mica. Group 4 was also statistically distinct from the Naxian cooking pots. The latter formed a very broad chemical group (Group 5) which could be divided into two (Groups 5a and 5b) in order to get lower standard deviations. In any case the archaeological conclusion remains the same: it is very unlikely that the cooking pots from Akrotiri were imported from Naxos. The texture of all cooking pots analysed was rather coarse which required the use of a larger-than-usual amount of sample (0.5 g) to ensure a representative analysis. Even so the results for Naxos show a very large spread of trace element compositions. This reflects the high overall variability in trace element concentrations of Naxian days.

The variability has been confirmed also by the analysis of Geometric pottery (Grimanis et al. 1989) and modern clays (Kilikoglou 1988). However the cooking pots from Akrotiri show reasonable standard deviations of their elemental compositions (Table 1), which also differentiates them from Naxos. The distinction between Group 4 and the two Naxian groups is demonstrated in the elemental bivariate plot of Fig. 3. Using the concentrations of Cr and Th, one can see that all the points corresponding to Group 4 are gathered together in a tight cluster, while the ones corresponding to Groups 5a and 5b are more spread out. In the same plot there is a slight overlap of the two Naxian groups, which results from the small number of specimens and the fact that this is not the ideal way to demonstrate a multivariate discrimination. In general, Group 5a is characterized by high Th and low Cr, and Group 5b by the opposite.

The Middle Cycladic sherds from Akrotiri had been stylistically divided into four groups: Cycladic white, bichrome ware, red burnished and dark burnished. Optical differences in the biscuits exist only between the burnished wares and the rest. The differences are the results of the special way in which the burnished pottery had been prepared and fired. This pottery is also slightly coarser, but no Middle Cycladic pottery analysed had any remarkable differences from the Late Cycladic local sherds and breasted ewers. The examination of all the Middle Cycladic analytical data did not show any obvious chemical discriminations among the various typological groups. They all seemed to form a large compositional unit which in terms of chemical compositions was very close to Group 1 and in many elements to Group 2. For this reason Groups 1 and 2 were considered as reference (core) groups and the multivariate probabilities of each Middle Cycladic specimen belonging to Group 1 and Group 2 were computed. Specimens having high probabilities of being members of one of the core groups and low of the other were clustered together. Two new Middle Cycladic groups were formed, one matching to Group 1 and the other matching to Group 2. As a result, seventy-two of the ninety-seven Middle Cycladic specimens clustered with Group 1, eleven with Group 2 and fourteen remained unclustered. The ones clustered with Group 1 included specimens from all four pottery classes, so there was no correlation between stylistic and chemical analysis results. Also the fact that most of the Middle Cycladic samples matched with Group 1 indicates that the same clay was used in both Middle and Late Cycladic periods. The eleven samples that clustered with Group 2 were probably imported from Phylakopi confirming in this way the contacts between the latter and Akrotiri.

The provenance of all the samples from Akrotiri that remained unclustered were possibly imported from other areas not included in this study. It could also be the result of natural variations in the clays or analytical errors, which should never be ignored.

A selection of nineteen Middle Cycladic sherds was examined under a scanning electron microscope (SEM), with an EDAX microanalyser attached to it. The examination basically included the estimation of the degree of vitrification in the microstructure of freshly fractured ceramic surfaces, the simultaneous chemical analysis of the body and the study of some black paint layers. As has been described in detail elsewhere (Maniatis and Tite 1981), the observation of the degree of vitrification, along with the type of clay, provides information on the firing temperature and the refractory properties of the clay used, factors which influence the ancient ceramic technology. The sherds examined in the SEM were representative of all four Middle Cycladic archaeological groups and the technological analysis results are listed in Table 2. As is the case with most Theran pottery (Noll 1978; Maniatis and Tite 1978; Aloupi and Maniatis 1990) the clay used is calcareous, with CaO concentration of the order of 16 ± 6%. However, its distribution is very inhomogeneous, giving rise to variable vitrification and microstructure. This behaviour is similar to Late Cycladic pottery (Aloupi and Maniatis 1990) and it is due to either a natural or an artificial mixture of a non-calcareous or slightly calcareous clay with large calcite particles.

From Table 2, it can be seen that the firing temperatures vary a lot, ranging from below 750° C to 1080° C. The most interesting feature, however, is the fact that the Cycladic white wares are fired predominantly at higher temperatures, indicating a different ceramic technology which distinguishes them from the rest. It is well known (Maniatis and Tite 1978; 1981; Noll 1987) that when calcareous clays are fired at temperatures above 850° C, and in particular between 850° C and 1050° C, they produce a cream-buff colour of consistent quality and with a high resistance to thermal and mechanical shock. On such a ceramic body a dark-painted decoration would also make a nice aesthetic contrast. These are exactly the features of the delicate class of Cycladic white pottery while the red and dark burnished wares, having a whole body coat and thicker walls, do not present a problem of contrast or durability and therefore there is no need for higher firing temperatures. An exception to this observation is the case of the two bichrome samples. One would normally expect them to be treated like the Cycladic white wares by the Theran potters, but their firing temperatures are lather low.

Examination of the dark decoration on the Cycladic white ware reveals that it is based on a mixture of Mn oxides, Fe oxides and argillaceous material. This kind of paint remains dark even in an oxidizing atmosphere and there is no need for a reducing-oxidizing cycle, but it sinters with difficulty, requiring high firing temperatures. Perhaps this might have been an extra reason for the high firing temperatures occurring in the Cycladic white ware. The paint of the whole body coats in the burnished ware is made of a clay suspension which in a reducing atmosphere would vitrify between 800° C and 900° C. The same is true of the red decoration of the bichrome wares. However, the black is made of Mn-rich material which stays black in an oxidizing atmosphere.

CONCLUSIONS

The breasted ewers from Akrotiri were probably made locally, though some of them (four out of twenty-nine) were imported from Phylakopi.

Only two out of ten stirrup jars from Akrotiri coincided chemically with stylistically similar ones from Knossos.

The cooking pots found at Akrotiri formed one group statistically distinct from the Naxian cooking pots, therefore it seems unlikely that they had been imported from Naxos.

Pottery from Akrotiri was present in Groups 2 and 3 coming from Phylakopi and Knossos, which indicates contacts among these sites and Akrotiri during the Late Cycladic period.

The majority of the Middle Cycladic pottery from Akrotiri had the same chemical pattern as the Late Cycladic control group, which indicates the use of the same clay in both periods. There were also contacts with Phylakopi.

The technological investigation of the sequence of Middle Cycladic pottery revealed that although the clay used for the body is more or less the same for all the groups, including refractory properties and basic raw material (as indicated by the neutron activation results), the procedure followed by the potters was different for the production of the Cycladic white from the rest. The difference is based on Mn paints used for the dark decoration of the Cycladic white and the use of high firing temperatures, while Fe paints and low temperatures in combination with reducing/oxidizing atmospheres are used for the red and dark burnished wares. The technology used for the bichrome needs further examination and experimentation to be fully understood.

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

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

Source: "Thera and the Aegean World III, Vol 1" (pp. 441-448)

Authors: V. Kilikoglou, C. Doumas, A. Papagiannopoulou, E. V. Sayre, Y. Maniatis and A. P. Grimanis 0