Petrographic Analysis of the Early Cycladic Wares from Akrotiri, Thera

The sherds came from the first and second destruction levels of buildings all over the site, and were analysed in thin section for purposes of both material characterization and provenance assessments.

Fourteen ware groups were represented, ranging from coarse storage fabrics to finer decorated wares thought to have been imported. The compositional uniformity of each ware group was determined in order to identify multiple fabrics which, in turn, could represent the existence of multiple workshops and/or sources of raw materials for each group. Of particular interest in this study were two very distinctive fabrics noted by excavators at other Cycladic sites with Early Bronze Age levels: Talc Ware, and a coarse fabric thought to contain glaucophane schist.

INTRODUCTION

Until very recently, the archaeological focus on the settlement at Akrotiri on Thera has been primarily on the Late Cycladic phase, its close associations with Minoan culture and its violent destruction. It has been obvious to excavators for some time, however, that the site was already settled before the middle of the third millennium BC (Sotirakopoulou 1986, 298 n. 1, 309). Though these early levels at Akrotiri have yet to be uncovered by systematic excavation, Early Cycladic sherds have been retrieved from the levels just above bedrock in the pits dug for the installation of roof pillars. A group of Early Cycladic sherds (most from the rock-cut structure in pillar pit 6) were recently published by Sotirakopoulou (1986), verifying the early occupation of the site and providing ceramic evidence for its relations with contemporary sites in the Aegean, particularly with those Cycladic sites of the 'Kastri' and 'Amorgos' cultures.

Since whole vessels of the Early Cycladic period at Akrotiri are extremely rare, studies of the pottery have depended on the survival of diagnostic decorative or morphological features on the sherds suggesting origins and stylistic affinities. As a further step toward understanding the Early Cycladic pottery at Akrotiri, and its reflections of cultural contact with other settlements, a programme of petrographic analysis of the wares was undertaken in collaboration with the excavators. The results of the study were used both to characterize the materials of the individual wares and fabrics (enhancing and refining the archaeological groups), and to provide a firm material basis for suggesting and/or corroborating sources for the pottery.

SAMPLES AND METHODOLOGY

Sixty-seven small samples of Early Cycladic sherds were selected from the first and second destructIon levels in excavated rooms from all areas of Akrotiri and submitted for examination as part of a wide-ranging programme of petrographic analysis of Bronze Age Cycladic pottery currently in progress at the Fitch Laboratory in Athens. The preliminary results of the thin section study are presented here and will be published subsequently in greater detail as part of the Cycladic project. Sherds were first examined in archaeological ware groups with a stereomagnifier in hand specimen on a fresh-cut surfare and the macroscopic features of the wares described using comparator charts (Munsell, Visual Percentage Estimation Chart adapted from Terry and Chilingar 1955: <0.5% very rare, 0.5-1% rare, 1-3% infrequent, 3-5% common, 5-10% frequent, 10-20% abundant, 20-30% very abundant, 30-40% dominant, 40-50% predominant; void shapes and sizes from Kemp 1985, 19). This was done to facilitate comparisons by other excavators, and to indicate the relationship between macroscopically and microscopically observable data.

A horizontal section was prepared for each sherd and identified only by a number for purposes of objectivity in determining material groups. The sections were studied with a polarizing microscope; estimations of inclusion shapes and percentages were made again using the same comparator charts and grain sizes were measured using the eyepiece micrometer. The macroscopic and petrographic results are presented by archaeological ware group, and then presented as material fabric groups (Table 1). Preliminary comments are made for each archaeological group of samples, and discussion of the overall results follows in the next section.

Table 1: Fabrics

MACROSCOPIC AND PETROGRAPHIC DATA

Twelve groups of wares, primarily identified as EC II-III, were sampled, some well defined in technical and stylistic terms, others comprising more loose-knit ceramic associations (e.g. Various Red-Pink). The number of samples taken to represent each group depended on the total number and size of sherds available and the degree of apparent stylistic and material variation within each group. All sherds are body sherds unless otherwise indicated, and sample numbers are those of the thin sections. The confidence of Sotirakopoulou's provenance assignments are indicated by a (?) where appropriate, and the strength of sample memberships in the groups are indicated by the author in the comments after the petrographic assessments.

     Dark-on-Light Painted Ware (local) 10-15:

     Hand specimen:      The ware is characterized by a thin dark paint (black to red-brown) often sintered, on a well-smoothed unslipped or possibly self-slipped very pale body and surface (10YR 8/3). The vessels were mostly jars and storage pots, with wall widths approx. 0.8 cm. Inclusion percentages vary from common to abundant (13 and 14 having the most, 10 having the least and finest grained), ranging from silt-sized to 2 mm, angular to subrounded, with rare vesicle voids. The inclusions consist of black and dark red volcanic rock fragments, white feldspars, red iron oxides, light brown (calcareous?) argillaceous rock fragments (ARFs) and small percentages of very fine mica.

     Thin section:      In thin section the group divides into two volcanic fabrics (found in other ware groups), hereafter referred to as Fabrics A and B. Fabric A, represented by 11, 13, and 15, is a calcareous fabric made from clays derived from pyroclastic volcanic sediments incorporating abundant percentages of carbonate material and well-preserved (often recrystallized or impregnated with red iron oxide) discrete calcareous bioclasts, including globigerinid, miliolid, coral, algae, endothyracid and occasional ostracod and brachiopod (?) shells (Fig. 1). The matrix is light brown and optically active, with inclusions frequent, poorly sorted (silt to 2 mm), angular to rounded. Silt-sized inclusions are common angular quartz grains, predominant recrystallized calcite, black iron oxides (magnetite?) and rutile, and rare white micas.

The larger inclusions consist of infrequent volcanic and argillaceous clasts, abundant carbonate rock fragments, common euhedral alkali (sanidine) and plagioclase feldspar (often weathered), biotites (0.25 mm) and iron oxides. Rare hornblendes and clinopyroxenes (augite) were also noted. The volcanic rock fragments (0.5-2 mm, rounded-subangular) were mostly porphyritic or microlitic rhyolite or dacite, occasionally andesite, spherulites common to the rhyolite, and the alkali and plagioclase feldspars occasionally zoned or containing glass inclusions. The groundmass of the glassy clasts is frequently perlitic, and some clasts show alteration (argillization or seriticization) and carbonate pseudomorphs of pyroxene. The argillaceous clasts have distinct boundaries and neutral density (except those which are slightly more ferruginous than the matrix), suggesting they were part of the original sediments or were formed during the preparation of the clay by the potter. The calcareous (micritic) clasts (angular subrounded), silt-sized to 2 mm, are often unevenly distributed throughout the matrix giving the fabric a very inhomogeneous appearance (cf. e.g. 'Milian bowls'). This feature was felt to reflect the quality of raw clay paste preparation rather than material selection since in 15 (and 24, Black Burnished, see below), a part of the sections has the more homogeneous appearance of the calcareous matrix of the Milian bowls fabric. The composition of Fabric A suggests the use of raw materials derived from the breakdown of pyroclastic and siliceous tuffite deposits which include abundant percentages of bioclastic limestones. Such formations exist within 100 metres of the site of Akrotiri, which were probably also the source for building stones identified in the settlement (Einfalt 1978, 461, 524). Such deposits also exist on Milos, where the tuffites and dacitic lavas are hydrothermally altered to bentonitic and kaolinitic clays. Fabric A has material parallels in later Theran and Milian wares studied by the author. Many Middle Cycladic Red Burnished and Black Burnished samples (Papagiannopoulou and Whitbread, unpublished) and Late Cycladic I pale Theran wares (Marthari and Vaughan, forthcoming) had similar compositions. Late Cycladic White Ware sherds from Phylakopi were also compatible with Fabric A, as were Middle Cycladic white wares examined from Mikri Vigla on Naxos: Late Cycladic White, Later Local, Fine Cycladic White (buff paste), and Dark Burnished (light fabric) (Vaughan 1989, 154-158). This volcanic fabric with carbonate inclusions was also compatible with Williams's Milian fabric group B from samples of later Middle Bronze Age pottery from Ayia Irini on Kea (Davis and Williams 1981, 296-297).

The second fabric, Fabric B, represented by 10, 12 and 14, was also derived from volcanic sediments, in this instance non-calcareous, non-fossiliferous deposits derived primarily from rhyolite and dacite lavas (see e.g. Fig. 2). The matrix of these samples was light grey-brown and optically active, with very abundant percentages of inclusions, primarily angular, some subrounded, poorly sorted silt to 2 mm sizes. The clay matrix was filled with frequent mesovesicle pores in a regular pattern resembling the distinctive 'web' appearance of some fired montmorillonite-rich clays (Vaughan 1987, vol. 2, 307). Silt-sized inclusions were common angular quartz and red and black (magnetite?) iron oxides, and rare white micas. Larger inclusions consisted of common percentages of porphyritic and microlitic volcanic rock fragments (0.25-2 mm) (tuff, rhyolite, andesite, glass) and argillaceous clasts (like Fabric A) (0.25-2 mm), round to subangular, and common discrete minerals, often relatively fresh and euhedral, occasionally large (0.75 mm): feldspars (predominantly alkali with some plagioclase), hornblende (often altered to red translucent anhedral grains), with some occurrences of augite, orthopyroxene (hypersthene) and very rare instances of olivine. The volcanic glass inclusions showed perlitic cracking, and in general the volcanic rock fragments were less altered than in the previous fabric, showing rare initial silicification or carbonate replacement, or alteration of the groundmass and pyroxenes to iron oxides. The slightly greater percentages and relative freshness of some of the volcanic rock fragments and discrete minerals suggest the pottery was made from primary clays resulting from the breakdown of these acid lavas. The composition of this group was similar to several Theran fabrics (XIX, XXI-XXII) identified by Pittinger (Jones 1986, 263), but at present has few material parallels with thin sections of later, pale Milian and Theran wares examined by the author.

     Black Burnished Ware (local) 23-27:

     Hand specimen:      The ware is characterized by a lustrous black sintered surface coat, over two types of fabric with walls approx. 0.7 cm thick; 24 and 27 have a pale fabric (10YR 7/1) with a common percentage of elongate pores aligned parallel to the vessel surfaces. Inclusions are rounded to subrounded, silt to 0.5 mm in size, and are calcareous (clasts, fossils) with lesser numbers of small dark volcanic rick and glass fragments. 23, 25 and 26 are fired to a darker brown colour (7.5YR 5/2) and have abundant, poorly sorted (silt-2 mm) angular to subrounded inclusions. These constituents are dark volcanic and pale tuff rock fragments, black pyroxenes, red and black iron oxides and feldspars. 23 was etched with a potter's mark 'T'.

     Thin section:      Thin section examination confirmed the distinction of the two fabrics, 24 and 27 being compatible with the calcareous Fabric A, 23, 25 and 26 belonging to Fabric B (Fig. 2). The carbonate constituents of the first group were consistently recrystallized, and in 27 the clay and fine calcite fractions were extremely well integrated giving the ground mass a very homogenous optical appearance. While the material composition of these samples was completely different from that of samples of contemporary Black Burnished Ware from Ayia Irini on Kea examined by the author, the calcareous samples were similar in composition to Middle Cycladic Black Burnished Ware from Akrotiri (Papagiannopoulou and Whitbread, unpublished).

     Red-Coated Incised Ware (local?) 19-22:

     Hand specimen:      This ware has a red (10R 5/8) surface coat, possibly once burnished, applied to a slightly uneven (unscraped) but smoothed surface. 21 retains traces of a dark surface layer (10YR 5/2) which may be representative of the original vessel colour or may represent a fire-clouded area of an otherwise oxidized surface since the fired colour of this sherd is also distinctive, being a very light grey colour (10YR 8-7/1), while the others are fired more brown (10YR 8/3). Inclusions are frequent, ranging in size from silt to 2 mm, angular to subrounded, with altered volcanic and argillaceous (reddish) rock fragments, light brown tuffs and argillaceous rock fragments, white feldspars and red iron oxides in 19, 20 and 23. 21 has black or partly altered volcanic rock fragments, whitish calcareous inclusions and clasts, and slightly more vesicle pores than the other sherds.

     Thin section:      The thin section data confirmed the existence of two fabrics in this group; 21, however, was paired with 19 in material terms, while 20 and 22 comprised the other type. The first two sherds belonged to the calcareous Fabric A, the carbonate in 19 being recrystallized, while in 21 it was finely dispersed as grains or present as rounded clasts (0.25-1 mm), with fossils showing very poor preservation. There was a slightly higher percentage of quartz silt in 21 as well. 20 and 22 were compatible with Fabric B, and were characterized further by higher percentages of matrix silt, the presence of ferruginous mudstone fragments (some silicified) and alteration (argillized) of the volcanic glass and rock fragments. Thus, while there were small constituent variations in 21, the different fired colours and percentages of pores noted in hand specimen were felt to be primarily the result of uneven firing conditions (e.g. atmosphere and temperature), probably random (e.g. due to location of the vessel and/or its access to oxygen) rather than any significant difference in material composition (though future studies might prove otherwise).

     White-Coated (local) 16:

     Hand specimen:      This ware was used for incised jars with wall widths approx. 1 cm. The surface showed abundant fine biotites and pale oxidized colours (10YR 8/2, 7/2). The sherd was unslipped but well scraped and smoothed, and the surface may have been whitened by the addition of kaolin, or soluble salts (Rice 1987, 336). Inclusions were present in common percentages: subangular-subrounded black volcanic and whitish calcareous clasts (0.25-2 mm), with some fossils visible, a few shale fragments, and red iron oxides.

     Thin section:      Thin section examination confirmed this sample's similarity to the other 'local' calcareous fabrics of type A. However, in addition to the usual constituents of such samples, this sherd also contained a few angular rock fragments (0.5 mm) exhibiting varying degrees of metamorphism: micaceous (muscovite) shale, quartz-biotite schist, ferruginous slate (foliated) with biotite and quartz inclusions distributed along bedding planes. While such clasts are still compatible with the formation of pyroclastic sediments which included fragments of metamorphic rock (in this case the crystalline basement complex of the calc-alkaline volcanic island arc), their presence was sufficiently unique to suggest the quarrying of raw materials from a different part of a pyroclastic deposit or perhaps from a different such deposit altogether.

     Milian Bowls 3, 4 and 5:

     Hand specimen:      This group had black-brown (7.5YR 4/2) or red (10YR 5/8) surface coats which, inthe case of the former, were slightly metallic-looking due to sintering. The red slip showed some reflective lustre along parallel pressure facets from burnishing. The body fabric is a pale grey-brown(10YR 7/2-3) and the bowl walls vary between 0.5-0.8 cm. The groundmass appears very fine-grained with abundant well-sorted, silt-sized flecks of biotite, few white feldspars and red iron oxides and tiny angular black inclusions, possibly glass (Fig. 3). There are a few larger, subrounded calcareous and volcanic rock fragments (1 mm) with some of the latter being partly altered to iron oxides. Pores are rare, occurring as microvesicles in the oxidized samples, and as microvughs in 3 which had a dark surface, possibly the result again of uneven firing conditions.

     Thin section:      The thin section data confirmed the material uniformity of this group which belonged to the calcareous Fabric A. These samples also all exhibited the extremely homogeneous calcareous matrix noted first in Black Burnished sample 27, and again felt to be the result of raw material preparation rather than source variations. This group was also characterized, however, by the presence of angular fragments of volcanic glass evenly distributed throughout the sherds, some showing slight chloritization or argillization along perlitic cracks. These fragments are more angular and represent relatively consistent sizes (0.3-0.5 cm, 0.75), though in 3 they appear only as the smaller size. The even distribution, consistent angularity and sizes, in contrast to the more usual variation in volcanic fabrics present in this fabric suggest this material may have been deliberately crushed and added as a temper for the manufacture of these bowls.

     Dark-on-Light Painted (Milian?) 6-9:

     Hand specimen:      The group was characterized by a black-brown sintered paint on an unslipped pale surface (10YR 8-7/2). The vessels represented were jugs and cups with walls approximately 0.5 cm. Two fabrics were apparent, the first (6-8) being Fabric A with frequent, poorly sorted (silt-2.5 mm) and unevenly distributed, subangular-subrounded inclusions. The dominant constituents are calcareous rock fragments and fossils, with a few reddish (altered) volcanic clasts. 9 represents a different very fine-grained non-calcareous fabric, with only rare ARFs (0.75 mm) present.

     Thin section:      The distinction between the two fabrics was confirmed by petrographic analysis with 6-8 being made of Fabric type A (Fig. 4). The inclusions were slightly more abundant and large at times: rare discrete crystals of augite (twinned), altered amphiboles and biotites (0.25-0.5 mm) with one fossil at 1 mm and a few ARFs 1-2 mm. 6 contained volcanic rock fragments with ferruginous or seriticized alterations suggesting a slightly more weathered deposit. 9 was similar to Fabric type B, though consistently fine-grained (only rare ARFs occurring of 1 mm) with the majority of rock and mineral inclusions being less than 0.3 mm, and fine biotites being present in slightly greater quantity than usual (possibly concentrated as a result of the raw paste's -refinement) (Fig. 5).

     Buff-Slipped Incised Ware (local?) 17 and 18:

     Hand specimen:      The sherds appear to have a very thin (self?) slip if any, fired the same pale colour as the body (10R 7-8/3). 17, from an incised jar, had walls 0.8 cm thick, while the wall of the other sample was half that width. Two fabrics were apparent: 17 had abundant poorly sorted (silt-2.5 mm), angular-rounded rock fragments (black-grey volcanic, pale calcareous), red iron oxides and black pyroxenes. 18 had a more consistently fine-grained and homogeneous groundmass with frequent (primarily silt-sized) subrounded inclusions, a few larger ones of 1 mm, porphyritic grey volcanic and reddish argillaceous rock fragments. Fine angular black glass (?) fragments and iron oxides were scattered evenly throughout the matrix.

     Thin section:      The thin section data confirmed the existence of two fabrics, that of 17 being compatible with Fabric A, while that of 18 being comparable to Fabric B with a slightly greater percentage of fine micas and vesicular glass fragments some showing initial argillization.

     Coarse Micaceous Ware (Imported) 28-34, 66-68:

     Hand specimen:      These sherds represented a heterogeneous group macroscopically, though all but 66 appeared very micaceous. They were comparable to the hearth fragments (4182, 4184 and 5795a) published by Sotirakopoulou (1986). Vessel types range from jars and bowls, to neckless globular jars and pithoid jars, with wall widths ranging from 0.6-2.5 cm. The oxidized fired colours were generally light red-browns (2.5YR 6-5/4-6) and incompletely oxidized cores ranged from greys (10YR 5/1-7.5YR 7/0) to weak red (2.5YR 5/2). 33 had a dark grey reduced surface (2.5YR 4/0), while 30, 32 and 66 had traces of a brown-black slip. The quality of surface preparation ranged from unsmoothed (uneven) (68) to smoothed with a soft tool only (29, 31, 34, 66) to well scraped and smoothed (28, 32, 33, 67). The primary inclusions were angular fragments of quartz mica schist, argillaceous and quartzitic rock fragments, quartz grains, fine black shiny minerals and discrete micas. However, these inclusion types showed notable degrees of variation within the group in terms of size, percentages and shapes. In some sherds the schist fragments (1-5 mm) were abundant (e.g. 29) while in others (e.g. 30, 32, 34) smaller (0.5-3 mm), less angular quartz rock fragments were most common. Biotite mica occurred in dominant to frequent percentages, ranging from silt-sized laths to discrete flakes of 1 mm, and in some samples (e.g. 28) appeared to be slightly altered. Mesovesicle pores occurred in few-abundant percentages.

33 and 66 were relatively distinctive in hand specimen. 33 (reduced) contained frequent subrounded quartz (0.5-1 mm) and a few red ARFs (0.5 mm) with a dominant percentage of silt-sized (rare 1 mm flakes) white micas, which were even more concentrated in the surface layer giving the finish a silvery reflective lustre. 66 contained a variety of subangular-round rock fragments: grey-green, black, red and light brown (calcareous?) with a few flakes of black and white micas. The matrix was dominated by elongate microvesicle pores, possibly the result of incipient vitrification.

     Thin section:      The matrix in the sherds was optically active, and inclusions were dominant (30-40%), consisting mainly of angular-sub angular schistose rock fragments, abundant angular quartz grains, K-feldspars (euhedral and anhedral) and occasional weathered volcanic or calcareous clasts (0.12-2 mm). Black and red iron oxides, rutile and fine micas were also present in varying percentages. Three somewhat heterogeneous fabrics were identified, the first included 29, 30, 33-34, 66. This group was characterized by frequent fragments of micaceous siltstones (0.25-2 mm), siliceous and micaceous (occasionally ferruginous) shales or slates with strong bedding planes, often foliated, with frequent discrete muscovite and biotite (altering to chlorite at times) (Fig. 6). The rock fragments in 33 were more schistose, with some sillimanite, more abundant discrete micas, and a few clasts of recrystallized carbonates. 66 had a similar profile (with slightly fewer micas), and was more fine-grained with inclusions not larger than 0.75 mm.

The second group of samples (28, 67-68) was characterized by the frequent presence of large (2.5 mm) subhedral feldspars, often zoned, predominantly alkali (sanidine, anorthoclase), frequent angular grains of quartz (0.25-0.5 mm), brown hornblendes, biotites altering to chlorite, and occasional, very weathered acid volcanic rock fragments (0.25-0.5 mm), and argillaceous siltstones (Fig. 7). In 67 and 68 the large K-feldspars showed varying degrees of alteration and some melt inclusions along zone horizons. They were frequently present in quartzitic fragments of schist, and discrete inclusions were rare microcline, microperthite, along with kyanite and chloritoid.

The third group (31, 32) contained frequent fragments of a quartz-mica-schist (muscovite), some micro-granodiorite clasts, frequent discrete clinopyroxenes (diopside) and orthopyroxenes along with large (0.5-1 mm) muscovites.

This collection of Coarse Micaceous samples was a somewhat heterogeneous group with material features associated with siliceous sediments from both schistose and granodiorite formations. Areas of contact between pelitic schists and granitic deposits compatible with this profile occur on the islands of Naxos, Mykonos, Tinos and Ikaria. The composition of this group overall resembles that of Williams's Group 3 Late Cycladic Theran sample, which he thought was not local in origin (Williams 1978, 509). In addition, some Middle Cycladic coarse micaceous wares from Mikri Vigla on Naxos, studied by the author, show very similar compositions to these sherds (Vaughan 1989, 150-159), and to several Naxian fabrics identified by Pittinger (Jones 1986, 263): her fabrics XIII and XIV are compatible with that of the first subgroup in this category (29, 30, 33-34, 66), her fabrics XV, XVI and XVIII are comparable to that of 28, 67-68, and her fabric XV to 31 and 32. Thus it is very probable that Naxos may have been the source for the fabrics in this ware group. The published hearth fragments (ibid.) had morphological features similar to examples from Delos and Kea, but the composition of the present samples was not similar to samples studied of coarse wares from Ayia Irini on Kea by the author (Vaughan, unpublished).

     Micaceous Ware (Imported) 35, 36:

     Hand-specimen:      The two sherds in this group were macroscopically somewhat different. 35 had a red slip over a well-smoothed surface containing abundant tiny biotites, and oxidized to a light brown (7.5YR 6/4) with a grey core with sharp horizons. The vessel wall was 0.8 cm thick, with abundant angular-subangular whitish rock fragments (0.5-1 mm) and mica evenly distributed.

36 had a surface appearance and feel similar to that of the Talc Ware sherds, with the same low (phyllitic?) reflective sheen, and red to greyish brown colours: 5YR 6/4, 10YR 5/2. The unslipped surface has been incised in a herringbone pattern. The sherd wall was 0.4 cm and contained abundant, poorly-sorted white, angular-subangular rock fragments (0.25-1.5 mm) and common micas.

     Thin section:      35, from a krater with a cylindrical neck and flaring lip, resembles 66 except for a dominant presence of fine (0.12 mm) biotites, oriented parallel to the surface, and occasional rounded-subrounded, often recrystallized clasts of carbonate, possibly added by the potter.

36 strongly resembles the fabric of Talc Ware (see below). The composition is dominated by angular fragments of an actinolite-chlorite-sillimanite rock, possibly originally a limestone. The fragments are poorly sorted, rounded to angular, and 0.12-2 mm in size.

     Talc Ware (Imported) 37-46:

     Hand specimen:      This ware, originally identified in the early levels at Ayia Irini on Kea (Caskey 1972, 373), and later studied by Wilson (1989, 42) is characterized by its distinctive soapy feel on relatively unsmoothed surfaces. The oxidized colours range from red-browns (2.5YR 4/4, 5YR 6/3) to light grey browns (10YR 6/2), and the surfaces are unslipped. Wall widths of the samples varied between 0.6-1.5 cm, and the fractures show a softened, rounded profile compatible with the smooth feel of the fabric. Inclusions are very abundant, consisting primarily of angular to subrounded grey, pink and white rock fragments with a micaceous, phyllitic appearance, primarily 0.25-2.5 cm, and evenly distributed throughout the fabric along with abundant flecks of (muscovite?) mica.

     Thin section:      Thin section examination confirmed the material homogeneity of this group. The rock fragments, present in dominant percentages, and ranging in size from 0.12-3 mm, were schist with chlorite, tremolite, plagioclase (albite?) and sericite, with occasional muscovite and talc inclusions. Quartzite clasts are rare, as are discrete clinopyroxenes (diopside?), rutiles and red and black iron oxides, and accessory garnets (Fig. 8). 38 has a small percentage of quartz grains (0.25 mm) and frequent clinopyroxenes (diopside). The groundmass consisted of the same minerals in finer form. The material profile of these samples is not compatible with local Theran manufacture, but rather reflects a source affected by relatively low-pressure metamorphism, possibly originally limestones (or in the case of 38, intermediate lavas). Such facies are common in the Cyclades and future fieldwork on Tinos, Syros, Siphnos, Amorgos and Ios (islands rich in such rocks) may provide more specific material parallels for such pottery. The composition of this distinctive ware found on Thera was not comparable to the composition of pottery samples published from Tinos by Etienne and Gautier (Jones 1986, 261), but it is identical to that of samples of it found on Keros, and in Periods II and III at Ayia Irini on Kea, where it was considered to be an import (Vaughan, unpublished).

     Glaucophane Schist Ware (Imported) 1, 2:

     Hand specimen:      This ware was unslipped but well-smoothed, and has a surface feel and low reflective sheen not unlike that of Talc Ware, except the fresh fracture has a sharper, more laminated profile. The oxidized colour is reddish-brown (2.5YR 2.5-4/4), wall widths average 0.5 cm and vessel types are unknown. Inclusions are frequent, poorly sorted (silt-3 mm), angular-subrounded, and unevenly distributed. The majority appear to be dark red-brown or grey-green phyllite or shale fragments, with a few quartz sand grains and angular black oxide inclusions scattered throughout the matrix. The metamorphic clasts are frequently micaceous (biotite), with quartz silt along the (often foliated) bedding planes. Pores are few and of mixed mesovugh-vesicle types, oriented parallel to the surface (as are the rock fragments).

     Thin section:     The matrix of the two samples was similar, optically active, with infrequent quartz silt and altered amphiboles, predominantly black iron oxides and biotites in frequent percentages. Sample 1 contained abundant angular phyllite and shale rock fragments (0.25-3 mm) with varying percentages of biotite, chlorite and quartz silt. The clasts are laminated by dark iron oxides frequently giving the fragments a stripy look. Occasional discrete feldspars are altering to sericite in places, and a few clasts of polycrystalline quartz and chert are present. This sample was somewhat reminiscent in composition to that of Williams's Late Cycladic buff-coloured Group 6 fabric, which he considered a probable import, possibly from Crete (1978, 511). However, it was also compatible with Pittinger's Amorgos fabrics XI and XII (Jones 1986, 263, 267).

2 contained a similar number of metamorphic clasts which appear to be derived from a glaucophane schist series, with inclusions of glaucophane and lawsonite and some epidote (Fig. 9). A small (0.2 mm) weathered volcanic rock fragment was noted, consisting of plagioclase laths in an altered groundmass. A large quartzite fragment was present (3 mm) and several shale fragments with biotites and muscovite, and the constituents showed strong parallel orientation to the sherd surfaces. These two samples may have originated from sediments from a single location, but the presence of glaucophane points to areas affected by high-pressure metamorphism such as exist on several islands in the Cyclades: Syros, Amorgos, Folegandros, Tinos and Siphnos (Seidel and Okrusch 1976, 349), and on Tinos and Seriphos these sediments are characterized by granitic intrusions (Durr et al., 1978, 469), compatible with the materials of these samples. Pittinger noted a rare ceramic fabric found on Milos however, containing glaucophane schist (Jones 1986, 263, 269), but it is thought unlikely that this island was the source for these samples.

     Various Pink-Red Ware Group (Imported) 47-56:

     Hand specimen:     This group was, by definition, heterogeneous in character, associated only by fabric colours and compositional appearances which differed significantly from the wares considered to be local. These samples were comparable to several (4183, 4185; 4189, 4190) published by Sotirakopoulou (1986) and attributed stylistically to the Kastri and Amorgos cultures respectively. The vessel types are unknown, wall widths varied between 0.6-1.3 cm, and some members were well scraped and smoothed with the possible addition of a thin slip (48-49, 51, 53-54), while the other samples were unslipped. Oxidized colours ranged from pink (5YR 7/3-4) to reds (10R 4/4, 5/6) to red-browns (5YR 5/4, 2.5YR 5/6). Some samples had grey cores, most with sharp firing horizons. There appeared to be three fabrics within the group: one large group, and two others represented by the single samples 47 and 54. The large group was characterized by abundant percentages of inclusions, either well sorted or bimodal (0.25 and 0.75 mm). The larger inclusions were subangular-subrounded quartz (grey, pink or yellow-white), feldspars, and the smaller inclusions were black shiny flecks. 47 contained a very abundant percentage of inclusions of three types: euhedral feldspars, quartz (2 mm), round-angular black volcanic rock fragments (0.5-1 mm) and rounded red ARFs (0.25 mm). Sample 54 had a dominant percentage of poorly-sorted (silt-3 mm) angular-subrounded siliceous and red ARFs, and red iron oxides, unevenly distributed.

     Thin section:      The macroscopic groups were confirmed as material groups by the thin section data. The clay matrix was in all cases optically active, with varying percentages of silt, oxides and fine laths (plagioclase?), and the total percentage of inclusions for each group was high - very abundant to dominant. The large group (48-53, 55-56) was characterized by abundant discrete, euhedral feldspars, predominantly plagioclase (oligoclase) exhibiting simple and multiple twins, discontinuous and convolute zoning, and showing a seriate size range (0.2-1 mm). Smaller (0.12-0.2 mm) angular quartz and dark iron oxides were also abundant, and twinned and zoned augite was common (Fig. 10). Rounded volcanic rock fragments were rare (0.25-0.5 mm) and of the rhyolite, dacite or andesite variety. 51 contained a few fragments of a glomeroporphyritic basalt, with plagioclase spherulites, augite and olivine. Discrete biotites and hornblendes were rare. The volcanic fragments in 55 were very altered to a mainly ferruginous clay groundmass.

47 was distinguished by an abundant percentage of volcanic rock fragments (andesite and rhyolite mainly), some with trachyte texture, some altering to sericite or clay in places. A few of these clasts appeared to be basalt (alkali olivine?), with carbonate pseudomorphs after olivine. Discrete subhedral alkali and plagioclase feldspars (0.2-0.4 mm) were abundant, occasionally zoned. Rounded siltstones (0.12-1 mm) were also abundant, while radiolarian mudstones, chert and apatites were present in rare amounts.

54 was dominated by poorly-sorted (0.2-1 mm), angular-subrounded clasts of clay and mudstones, occasionally partly silicified, with rare red iron oxides. Small fragments (0.2 mm) of rhyolite were rare, as were clasts of chert, and inclusions of quartz, alkali feldspar and altered amphiboles. Iron oxides were abundant throughout the matrix.

These samples represent altogether ferruginous sediments derived from volcanic sources of a primarily acid-intermediate nature, with 51 and 47 containing a few clasts of a more basic volcanic type. Their composition suggests a non-local source, and while some hornblende-rich diabase outcrops on the nearby island of Anaphi (see also Einfalt 1978, 466), the origins of the group may lie outside the immediate volcanic arc of the Cyclades, possibly in Aegina or farther afield. The author does not discount a possible origin for this material on Syros or Amorgos (following the stylistic attributions), but fieldwork and further ceramic analysis is required to determine the extent to which such volcanic sediments occur on these islands.

     Fine Wares (Imported) 57-65:

     Hand specimen:      This group of samples was rather heterogeneous in nature, though consistently fine-grained, very compact and well-smoothed and finished. 57, 60, 64-65 showed some macroscopic similarities, and the others were more distinct. 57 had a shiny (not burnished) black slipped surface (5YR 2.5/1) over a light brown body (10YR 6/2) and grey core. The rim fragment was 0.4 cm in width with common elongate mesovesicle pores aligned parallel to the surface, and silt-sized inclusions. A few rare yellow or white clasts were visible (0.12 mm).

60 had slightly mottled surfaces and showed areas of surface sintering, with slightly more argillaceous-looking clasts, but otherwise similar to 57, as was 65.

64 had red paint in a criss-cross design, slightly reduced in areas, and in general the surface looked somewhat light grey due to overfiring (or possible exposure to secondary heating at some point). A thin oxidized body horizon in the 0.4 cm wall showed light brown colours (7.5YR 6/4), while surface colours were more grey (10YR 6/2, 7/1). Abundant mesovesicles showed orientation parallel to the sherd surfaces, and the only inclusions visible were abundant tiny biotites (?) on the surface.

58 had a slightly mottled slip, with colours ranging from pale red-brown to red-grey (5YR 6/4, 5/2). There were frequent white inclusions, angular-subangular, 0.25-1 mm, a few red-orange ARFs (0.5-1 mm) and some rare muscovite (?) and mesovesicle pores.

63 was extremely fine-grained and compact, 0.4 cm in width, and decorated with thin parallel lines of matt brown paint on a slightly pinkish surface (7.5YR 8/4). The surface was unslipped and showed imperfect (streaky) burnishing, and the fabric contained very rare, angular white inclusions (0.25-0.5 cm), a few silt-sized red inclusions and almost no pores.

59 was very unusual: the body was oxidized (5YR 6/4), but was covered with a matt white substance, which was, in turn, covered by a yellow (10YR 8/4) burnished slip. It is thought unlikely that the white coat represents any kind of adhesive, but rather was applied perhaps to lighten the surface to which the pale slip was applied. Abundant subround-round silt-1 mm inclusions were noted (black and brown), rare mesovesicle pores and some extremely fine mica on the surface.

61 was incised in a herring-bone pattern on a matt, unslipped mottled surface: 5YR 6/3 to 5YR 4/1. Abundant round-angular clear and white (quartz?) inclusions were noted, common subrounded black volcanic rock fragments (0.25-1 mm) and smaller red argillaceous clasts, with abundant fine biotite visible on the surface.

     Thin section (62 was lost in preparation):      Thin section examination confirmed the heterogeneous character of this fine-grained group, with most sherds having unique fabrics, apart from 57, 64 and 65 which were very similar. The clay matrix of these three samples was optically inactive and included abundant brown iron oxides, angular quartz silt, biotites and pale laths (plagioclase or sericite?) which, with the micas, showed parallel orientation to the surface. Larger inclusions were only abundant quartz (0.15 mm) and occasional chert clasts (0.25 mm), with rare round ferruginous ARFs. The constituent profile of this siliceous group was insufficiently distinctive to suggest an area of origin. This group was slightly reminiscent of Williams's Late Cycladic Group 7 fabric from Akrotiri, thought to be a Cretan import (1978, 512). More significantly perhaps, these samples showed strong material similarities to thin sections of fine imported wares from Period II at Ayia Irini on Kea: Dark-on-Light Painted and Yellow Mottled sherds from sauceboats and pedestalled cups (Vaughan, unpublished).

60 was also materially undiagnostic. The matrix was optically active, containing quartz silt, calcite grains, biotites, abundant black iron oxides and frequent fine white laths (plagioclase?). The larger inclusions were very abundant angular quartz (0.2 mm), a few random carbonate clasts (0.3-1.5 mm), fine biotite and occasional chert, fine weathered plagioclase laths, rare fine muscovite, and a few rounded ARFs (0.75 mm) with rare quartz inclusions.

58 had an optically active matrix containing quartz silt, fine biotites, iron oxides and fine plagioclase laths. The larger inclusions were distinctive, derived from a predominantly phyllitic source. Abundant subangular-subround (0.25-1.25) micaceous (biotite) phyllite fragments with varying percentages of quartz (silt, grains and chert) as discrete inclusions, lenses, or as a replacement fabric, dominate the composition. Some of the clasts show brown iron oxide laminations (occasionally cross-laminations). Rare andesitic rock fragments are present (0.25-0.5 mm) with an altered groundmass and traces of pyroxenes or hornblendes. A few discrete subhedral alkali feldpsars (0.12-0.25) and biotites (0.5 mm) also occur (Fig. 11). The source of the iron-rich sediments for this sample was probably an area of low-grade metamorphism in contact with fine-grained volcanic deposits, such as outcrop frequently in the Cyclades and elsewhere in the Aegean, though not on Thera.

63 had a related material profile to that of 58, in that its materials reflected a source affected by low-grade metamorphism. The matrix was optically active, and contained very fine biotite, oxides, silica and amphiboles in equal and common percentages. The larger inclusions were primarily infrequent, subrounded fragments of equigranular siltstones, most 0.5 mm with a few at 2 mm. They were composed of amphiboles, fine angular quartz, biotites and red iron oxides in an argillaceous matrix. There were also present a few discrete biotites and quartz grains (0.05-0.1 mm) and pale laths (plagioclase?) in the sherd matrix (Fig. 12). This fabric bears a strong material resemblance to a Middle Cycladic fabric found at Mikri Vigla on Naxos, studied by the author (Vaughan 1989) whose origins were suggested to be in central Crete where there are extensive phyllite formations (Seidel and Okrusch 1976, 348; and personal communication from P.M. Day). This sample was also similar to the Late Cycladic fabric described by Williams for Group 10 from Akrotiri, thought unlikely to be local (1978, 513).

59 and 61 were more volcanic in nature. The first had an optically active matrix with rare quartz silt, abundant biotites (oriented parallel to the surface), and iron oxides. The larger inclusions were frequent angular-subangular fragments (0.12-0.5 mm) of andesite/dacite rocks. Some of these clasts were glassy with perlitic cracks, or porphyritic with plagioclase glomerocrysts or rare microperthite. Discrete brown oxides (0.25 mm) were frequent and biotites (0.25 mm) and alkali and plagioclase feldspars (0.4 mm) were slightly less common. A few round mudstone clasts (0.25-0.5 mm) were also noted. While more ferruginous than the pale-firing volcanic fabrics previously attributed to either Theran or Milian sources, this fabric also represented sediments with fine-grained acid volcanic constituents. These constituents may have been added by the potter as temper in this case (further sampling would be required to establish this point), though considered unlikely since the matrix and discrete mineral inclusions were compatible with a volcanic source. 61 was very similar to the non-calcareous Fabric B of the pale wares (e.g. Black Burnished 23, 25-26),with the exception of the presence of angular mudstone clasts (0.25-0.75 mm) in common percentages, and greater quantities of iron in the clay resulting in redder fired colours. Therefore a non-Theran (and non-Milian) source is suggested, though such materials are compatible with many other areas of the Cyclades.

DISCUSSION

The ceramic fabrics of the archaeological groups represent a variety of raw materials and sources (Table 1), ranging from clays derived from fine-grained pyroclastic/tuffite and acid lavas, to micaceous-phyllitic and schistose clays, siliceous sediments, and finer-grained, less distinctive sedimentary fabrics. In the absence of an existing reconstruction of the geological profile of Thera prior to the Late Bronze Age eruption, suggestions of material provenance for samples thought to be local must remain somewhat speculative. For example, it is possible red-firing clays were once available from weathered exposures of local phyllite or lavas, which may have been used to produce redder-firing, and possibly finer-grained wares with small inclusions of the parent rock.

There is some proof of this idea since a fine-grained clay was discovered in some quantity in a rock-cut structure in pillar pit 17 (Sotirakopoulou 1986, 298). It was not known whether this clay was the product of a natural silting process, or whether it represented a deliberate deposit for use in ceramic manufacture at the settlement (C. Doumas, personal communication). A sample of this clay was taken by the author and made into briquettes, fired at both 800° C and 1100° C. There were no inclusions visible in hand specimen, and the oxidized colour was a yellow-red (5YR 6/6). At 1100° C, however, the clay became completely vitrified, and had a slightly metallic sheen, with a much darker, more purple-brown colour (10R 3/4). Petrographic examination of the briquettes confirmed the extremely fine-grained nature of the clay: the majority of inclusions were silt-sized, with rare larger inclusions (0.25 mm). Common oxides, and abundant biotites and fine plagioclase laths characterized the clay matrix, with rare quartz silt. Larger, discrete, anhedral weathered plagioclase were few, as were hornblendes, polycrystalline quartz clasts, and very weathered, porphyritic andesitic rock fragments, with phenocrysts of hornblende, biotite and plagioclase in an altered groundmass.

The importance of this comparative fired clay sample is both material and stylistic in nature. The composition of the clay is compatible with the local lavas, and therefore suggests such red-firing, and fine-grained raw materials were available to local ancient potters. This fired clay bore a strong petrographic similarity to the compositions of Fine Imports 57 and 65, though they contained larger quantities of quartz silt (added?) and were fired to a more grey-brown colour. In addition to this clay being available for the production of local, redder fine wares at Akrotiri, it may also have provided the base for some red or purple-brown paints used on the local decorated pale wares.

It seems probable that metamorphic exposures, however, were limited (as they are today on both Thera and Milos), and that potters would have preferred producing the pale wares, as they constituted a much more distinctive product in a ceramic market dominated by redder fabrics made from the clays derived from the extensive Cycladic metamorphic formations on other islands. In the current study then, the material evidence would seem to corroborate Sotirakopoulou's assessments as to local or imported origins for these Early Cycladic wares.

The question of the provenance of the pale wares remains a complicated and interesting problem however. Weathering of the pyroclastic, tuffite and lava formations near Akrotiri probably provided the source of the clays for the local pale wares. Einfalt noted the alteration to kaolinite of the lavas on Nea Kameni and the dacitic tuffs on Akrotiri peninsula (1978, 461). However pale-firing clays with virtually the same material origins are (and would have been) available on Milos. Since the success of provenance assignments on the basis of material compositions of pottery depends upon the relative uniqueness of the constituents of the fabrics and the sediments of the potential source areas, distinguishing between the volcanic clays of Milos and Thera would be exceedingly difficult, as previous chemical work has already demonstrated (Jones 1986, 280). There have been two sets of hopeful results in relation to this problem however: the first was a limited heavy mineral study published by Williams which suggested that the presence of frequent biotites and hornblendes characterized the Milian samples and were absent in the Theran sherds (1980). The other set of results, from NAA of MC-LC I pale wares considered local to both Phylakopi and Akrotiri, found that the elements Sm, Yb and Lu were successful in discriminating between the ceramic products of the two islands (Kilikoglou 1988). Reproduction of these results with well-defined sample groups is desirable if the analytical success is to be translated into a reliable approach to the problem for other scholars.

The difficulty in making such a distinction on the basis of petrographic analysis, however, is highlighted in the current data by the presence of both Milian and Theran wares in each of the Fabrics A and B in this study. Thus it was felt that the focus of inquiry must be on understanding the two fabrics in technological as well as material terms, rather than trying to distinguish between origins on Milos or Thera for the present.

It is possible the two fabrics represent different contemporary traditions, or a chronological development during the Early Cycladic period in production of the pale wares, but further sampling from stratified deposits of comparable wares from both Akrotiri and Phylakopi would be required to test this theory, especially since Pittinger's Theran fabrics include no reference to carbonate constituents (Jones 1986, 263). This is curious since examination of thin sections of later pale wares from both sites by the author indicated a definite preference by the potters for volcanic clays containing varying types and percentages of carbonate inclusions. Since some of these wares had darkened or red surfaces (e.g. the Middle Cycladic Black or Red Burnished Wares) it was obvious the local potters preferred to continue to use traditional, pale-firing clays and add pigment to the slips than to try using more ferruginous local clays, which might have produced those surface colours without added pigment. It may be that red-firing volcanic clays were (and are) extremely rare on both islands (though the clay from pillar pit 17 would tend to discount this), or that the pale ceramic fabrics were simply a more distinctive and successful product; fieldwork may clarify the degree to which such clays were an option to the local potters on Thera and Milos. Pittinger suggested the lava-derived clays were associated with the darker ceramic fabrics on Milos (Jones 1986, 263, 269), and this would be compatible with the evidence from the experimental briquette from Akrotiri. However the non-calcareous volcanic (lava) clays in this study were also found to be present in samples of pale fabrics, and rare samples of fine-grained Cycladic White from Phylakopi.

The carbonates in Fabric A of this study are not considered to represent the addition of temper by the potter for a number of reasons: the carbonate content in each ware, and sometimes sample, varies substantially in type, relative abundance, size and distribution; calcite is also present as a constituent in the groundmass of the sherds; the presence and preservation of the discrete fossils seems incompatible with crushing of the carbonate for temper; the age of the fossils is incompatible with the locally-available Triassic limestone on Thera.

Einfalt also noted in thin sections of several Late Cycladic Theran sherds the presence of Eocene microfauna (1978, 461), and suggested their origin may have been in the carbonatic layers of the phyllite rocks of Thera, which are of compatible age. The present data would seem to argue against this since occurrences of fragments of metamorphic rocks in the pale volcanic fabrics are extremely rare (apart from Whitish Coated sample 16). Einfalt also suggested, however, that the fossils may have survived in xenoliths in the dacite pumice tuffs or pyroclastic outcrops on the Akrotiri peninsula (1978, 467-468). The pyroclastic sediments near Akrotiri contain intercalations of marine silts and microfossils, and are locally altered to bentonite and kaolinite clays (Dr Y. Kanaris, IGME, personal communication). Such intercalations are also present in the young tuffitic, or Neogene tuffites on Milos (Dr M. Fytikas, IGME, personal communication). Thus the carbonate inclusions and fossils almost certainly originated as natural constituents of the parent rocks of the volcanic clays. The fact that Einfalt noted phyllitic rock fragments in the Late Cycladic Theran sherds, as did Williams (1978, 508), may point to a chronological change in quarrying location for the volcanic clays, or their deliberate inclusion by the later potters. This idea is reinforced by the fact that metamorphic rock fragments were not noted by Williams in thin sections of later Middle Cycladic pale wares from Ayia Irini on Kea, which he assigned to a Milian origin (Davis and Williams 1981, 296-297).

In the light of the material similarity then of the volcanic clays from Milos and Thera several remarks can be made. Fieldwork dedicated to discovering the extent to which sediments from the two islands can be distinguished is required, and is, indeed, in progress (Vaughan, unpublished). In addition, it is hoped that the same research will be able to indicate the degree to which the tuffite and dacitic lava deposits are naturally varied with reference to inclusions of carbonates, fossils and metamorphic rocks, either regionally or in vertical section. This knowledge would help resolve the question of whether the two volcanic fabrics (A and B) represented the use of a single type of clay with significant natural compositional variations, or whether the distinction indicates a more radical difference in quarrying location (regional, ot vertical within a deposit) possibly reflecting chronological differences in the archaeological samples. Until such information is available, however, the weight of provenance assignment, for these pale wares in particular, must still rest with the archaeologist. This responsibility increases the requirement in such cases for high-quality stylistic and morphological ceramic data, preferably known from whole vessels and found to be regionally specific, on which to base the provenance assessments.

A few general technological points can be made from the study. The vast majority of the Early Cycladic wares examined represented pottery fired to temperatures not greater than 850° C, leaving the clay matrix optically active in most cases. The few exceptions were the extremely fine-grained imported fabrics which reflected different material and manufacturing traditions altogether. In addition, the majority of the fabrics appeared to represent the use of raw materials prepared without the addition of tempering materials. The few suggested exceptions were the Milian Bowls (added crushed volcanic glass?), Micaceous sample 35 (added carbonate?) and Fine Ware import sample 59 (added crushed volcanic rock?).

CONCLUSIONS

The Early Cycladic wares at Akrotiri represent a sophisticated and wide-ranging ceramic assemblage, and in material terms, confirm Sotirakopoulou's suggestions of contact with other contemporary cultures, both in the immediate Cycladic region, and probably farther afield (1986, 309-310). It was impossible to distinguish by petrographic analysis either the local Theran, or possible Milian, origin of the pale wares, and so for the purposes of this study the archaeologist's provenance assignments were provisionally accepted, pending the results of future fieldwork and additional Early Cycladic ceramic finds from Akrotiri. Thus the local pale wares consisted of Painted Dark-on-Light, Black Burnished, Red-Coated Incised, Whitish-Coated and Buff-Slipped Incised, while the Milian Bowls and Milian Dark-on-Light samples were left as imports from that island.

The Coarse Micaceous Ware and Micaceous Ware samples pointed to contacts with Naxos, while the Talc Ware and Glaucophane Schist Ware (2) strongly reinforced Sotirakopoulou's suggested contacts with cultures on such islands as Amorgos or Syros, though other Cycladic islands also offer compatible material sources but are less well known archaeologically: Ios, Tinos, Folegandros and Siphnos. Some of the Various Pink-Red Wares and Fine Ware imports suggested contacts with Crete, and possibly other islands such as Anaphi or Aegina.

More specific and useful comparisons with previously published petrographic analyses of ceramics from the Cyclades and Thera were often hampered by the relative absence in those reports of precise, detailed compositional descriptions, though these studies are acknowledged to be helpful, initial contributions to the field. It is hoped that the wide-ranging petrographic study of Bronze Age Cycladic fabrics currently in progress, of which this report is a preliminary part, will continue to elucidate the complex relations between Cycladic cultures, already well established it seems, in the Early Bronze Age period.

Addendum

During the course of investigations of ancient Milian and Theran pottery and their raw materials, several discoveries were made (subsequent to the submission of the original paper) which shed light on some of the problems discussed, particularfy those to do with distinguishing the origins and characterizing the technology of the Cycladic pale wares common to both islands throughout the Bronze Age.

The evidence for the existence of a non-calcareous volcanic fabric for some Cycladic pale wares set out in the paper was very important as it was the first time such clays have been identified with the manufacture at Akrotiri of wares other than coarse or cooking wares. The petrographic evidence was double-checked by chemical analysis of the samples, and the calcium oxide content was confirmed to be significantly lower than 5% (R.E. Jones, personal communication).

The calcareous Fabric A samples were also re-examined petrographically as a group and compared to thin sections of later pale wares from Milos and Thera. All but two of thirty-eight Middle Cycladic Theran Dark-on Light, Bichrome and Burnished Wares belonged to Fabric A, while only two of ten Late Cycladic Theran pale ware samples were the non-calcareous Fabric B (fragments of a pithos and of a tripod vessel). A different pattern emerged, however, when a small group of Middle and Late Cycladic pale wares from Phylakopi were examined. The fabrics of Bichrome and Cycladic White Wares were consistently Fabric A, though the carbonate percentages varied greatly and microfossils were not always present. Burnished Wares, however, were predominantly non-calcareous, unlike the Theran samples. In addition, all Middle and Late Cycladic pale ware imports studied from Ayia Irini on Kea, and from Mikri Vigla on Naxos, belonged to calcareous Fabric A. The existence of a non-calcareous material tradition for some Cycladic pale wares then was confirmed, though it was apparently most common during the Early Cycladic period at Akrotiri.

The origins and preparations of the raw materials for Fabrics A and B were still problematic, however. The non-calcareous volcanic fabric was compatible with sediments derived from either Milian or Theran lavas, but experimental work on such clays has confirmed their red-firing character. Thus the pale-fired colours of Fabric B must have been achieved by the addition of another clay, probably kaolinite, common on both islands.

The question of the source of the microfossils in Fabric A, thought to be related to the type of clay deposit, was left unanswered after extensive sampling and study of Milian clays failed to uncover any clays which were characterized by such constituents, even those clays derived from the pyroclastic and tuff formations. However, a lithic sample from a young tuffitic formation just south of the site of Phylakopi on Milos suggested a possible solution. This sample contained both carbonates (and microfossils) and fine volcanic rock fragments comparable in type and percentage to those characterizing Fabric A. There were no clay beds noted in association with this formation at the top of a hill, but the rock was very weathered and friable and would have been easily quarried and crushed as tempering material for the nearby abundant non-calcareous alluvial clays. The addition of such a calcareous component in limited quantities to a non-calcareous volcanic clay would have served two purposes: the calcium would have increased the refractory property of the paste while also ensuring paler fired colours. The inhomogeneous optical appearance of the matrix of some thin sections of Fabric A may reflect such a mixture of raw materials, and further study of the samples along with replication experiments are hoped to provide more information on this point.

Although the present evidence comes from Milian clays and associated rock formations, the geological similarity of the tuff and pyroclastic deposits on Milos and Thera made theoretical parallels plausible. Fieldwork is continuing on Thera to locate any sediments which may provide similar material evidence for a local source for the raw materials for Fabric A, though with so few exposures of ancient deposits, most discussion may have to remain speculative in nature. It can be said, though, that the topography of Milos is (and most probably always has been) more benign than that of Thera with respect to the formation and accumulation of clays. The geothermal activity on Milos (promoting the alteration of rocks to clays) is also more extensive than that on Thera, and thus the enormous quantities of fine quality clays on Milos have probably always exceeded the amounts of similar materials available on Thera. It is possible consideration should be given to the idea that the bulk of pale Cycladic wares was manufactured on Milos, with perhaps some being imported to Thera where they were decorated in the local koine. Perhaps such problems will provide new impetus to the excavation of the important industrial peripheries of Akrotiri on Thera, where the ceramic installations and workshops might be found, with the evidence required to reduce the element of speculation in investigations of archaeological ceramics from the site.

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

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