Spacial and Temporal Uniformitarianism in Late Cycladic I: Perspectives from Kea and Milos on the Prehistory of Akrotiri

We aim to show that sensitivity to spatial or temporal variation in the data can improve our appreciation of the position of Thera in the wider Aegean world. Two topics are examined in some detail as examples of this general theme.

First, notwithstanding the obvious difficulties to be faced in describing the cultural landscape of Thera at the time of the eruption, it has become clear from intensive surface surveys conducted on Milos and Kea that there existed major differences between the distributional patterns of settlements on these two islands and that on Thera. In particular, the population of Thera was probably much larger, the range of activities conducted outside the site of Akrotiri was far wider than at Ayia Irini or Phylakopi, and the relative degree of settlement nucleation and site hierarchy implies different strategies of exploitation and control of these sites' hinterlands (in the Theran case, perhaps more 'Minoanized').

Second, freshly completed studies of the stratigraphy of Ayia Irini in Period VI make it possible to describe the development of ceramic styles there within the Late Cycladic I period, and to compare them with the chronological distinctions recently proposed by Marthari for Akrotiri and by Cherry and Davis for Phylakopi. It will be argued that evidence from both Ayia Irini and Phylakopi demonstrates that the destruction of Akrotiri occurred some time before the end of LC I or LM IA, thus weakening the suggestion that the Theran catastrophe happened at a time when the LM IB style was already current in Crete; but it also illustrates the shaky basis of attempts to write an 'event-based' history from the suppositious correlation of deposits that appear stylistically similar.

INTRODUCTION

Aegean prehistorians sometimes seem curiously uncritical in their ready assumption that the history and pattern of settlement at different sites or on different islands were very similar - in some cases despite clear evidence to the contrary. Contrasts and variations between their archaeological records are very often ascribed to the serendipitous aspects of site discovery, to the vagaries of preservation, or simply to lack of information as a result of insufficiently rigorous investigations; and sometimes, of course, these are the salient factors. The alternative possibility - that observed differences are genuine and that, in the right circumstances, the absence of finds can be just as informative as their presence - is often explored only as a last resort. Yet why should it be expected a priori, for instance, that there was a widespread 'typical Late Cycladic settlement pattern', or that the sequence of stylistic change revealed in the pottery from one or two key sites should stand proxy for trends in the region as a whole, or even in individual parts of it? Uniformitarian assumptions of this sort, both spatial and chronological, seem to us dangerous, in that they discourage the exploration of the very variability which might open the way to the solution of problems of great interest.

As exemplifications of this general theme, we consider in this paper two case studies based on our work (both mutual and individual) in the islands of Kea and Milos, and particularly on the results of a new survey of north-west Kea and of recently completed studies of the stratigraphy of Ayia Irini in Period VI. In both instances we aim to show that sensitivity to variation in the data yields results that enable us to appreciate the position of Thera in the Aegean world with improved precision and greater insight.

(I)      It has often been suggested that the material culture of Thera was affected much more than that of any other Cycladic island as the result of its contacts with the Minoans. The town at Akrotiri, especially, has been interpreted variously as a Minoan colony, as an overseas possession that was governed from a palatial centre on Crete, or as the seat of a fully autonomous Cycladic polity which had adopted Cretan methods of administration and control (Renfrew 1978, 418-420). Ayia Irini and Phylakopi, on the other hand, while obviously thoroughly 'Minoanized' settlements, seem less so when set beside Akrotiri. We argue here that differences between these centres are mirrored by related contrasts in the organization of their hinterlands, and that these contrasts are not merely the result of differing conditions of preservation, or of accidents of discovery, or of the history of field research in each island. Despite the unique difficulties of studying the Late Bronze Age landscape of Thera, there is growing evidence from the island for a functionally and hierarchically differentiated system of settlement subsidiary to the unusually large and complex site of Akrotiri. This pattern is wholly unparalleled on Kea or Milos, where careful surveys have been conducted, or on other Aegean islands, and we consider whether it may reflect a Minoan rather than a Cycladic form of political economy.

(2)      The settlement history of Akrotiri, as it is presently understood, is punctuated by two major destructions: the first not long after the beginning of the Late Bronze Age, the second at the time of the final abandonment of the site. At Ayia Irini there was also a destruction within the LC I period, and both there and at Phylakopi it has been possible to describe the development of ceramic styles within the period - in the former case using selected deposits, in the latter a more continuous and quantifiable sequence of pottery from stratified levels. We briefly summarize the LC I ceramic sequence at Ayia Irini, compare it with those of Phylakopi and Akrotiri, and ask whether new information from Kea and Milos can be of use in sorting out the chronology of the abandonment of Akrotiri and the volcanic eruption that buried it.

PATTERNS OF SETTLEMENT ON KEA, MILOS AND THERA IN LC I

The history of the settlement at Akrotiri can scarcely be understood except in its regional setting: what was the nature of its relationships with other centres of the Aegean Bronze Age, and how did they affect its growth and prosperity? It has been argued in several recent papers (e.g., Cherry and Davis 1982; Schofield 1982) that Thera, together with Milos and Kea, served as stepping-stones along one of the major sailing routes between Crete and the Greek mainland, the so-called 'Western String'. In the Middle and Late Cycladic periods, both Cretan and mainland pottery reached these centres, while exchange of ceramics between them has also been documented. Copper, silver and lead from the mines of eastern Attica could explain, at least in part, the specific attractions of Kea (Gale et al. 1984), although this island may also have possessed its own ores (Caskey et al. 1988). Why was Minoan pottery so widely imitated, and why were technological innovations diffused from Crete? Many studies (e.g., papers in Hägg and Marinatos 1984) have considered possible relationships between changes in the material culture of these sites and their evolving political and economic circumstances.

In the decade since the last Thera and the Aegean World congress, our picture of the Cyclades during LC I has changed very little in outline. The discovery and careful exploration of contemporary settlements in the central, northern, and eastern Cyclades still remains a major desideratum, though fresh investigations on Naxos are finally yielding new Minoan finds (Barber 1987, 72, 162, 189). Yet merely to document such contact is insufficient: in order to evaluate the impact on local developments of contact and exchange with Crete, and to discriminate between various existing models of Cretan-Cycladic exchange, thorough large-scale excavations of settlements in these areas are needed. Nonetheless, the suggestion that the islands of the southern and western Cyclades enjoyed a 'special relationship' with Crete, or that their centres were in more frequent contact with the Minoans than those of the central and northern Aegean, has not been undermined by recent discoveries and research. On the other hand, the completion of intensive surface surveys on Milos (Cherry 1982a) and, more recently, on Kea (Cherry et al. in press) has provided significant information of a new kind about the organization of settlement at the local level. It is possible, at last, to describe the relationships between certain Cycladic centres and their individual island territories, and thus to contrast Theran settlement patterns with those of other islands in more detail than was possible when Wagstaff attempted to do so over ten years ago (Wagstaff 1978). The interest of such an exercise lies in the fact, now more clearly demonstrable, that not all LC I centres interacted with, or controlled, their hinterlands in the same fashion.

The results of systematic survey on Milos in 1976-77 have been described elsewhere (Cherry 1981; 1982a) and need not be repeated in detail here. When taken in conjunction with that of three earlier one-man surveys and decades of casual exploration and chance discovery, the evidence suggested clearly that the EC pattern of widely scattered small habitation sites and their associated cemeteries gave way to a far more highly nucleated settlement system which persisted throughout most of the second millennium (Cherry 1982a, Fig. 2.3, 2.4). Indeed, other than Phylakopi itself, the only documented MC location is at Kapari (Cherry 1982b, 296), which is so close to Phylakopi as to be perhaps almost a part of it; for early LC times, not a single find is known, despite the evidence of very considerable activity during this period at Phylakopi (Renfrew 1978). The natural inference drawn from these data by the various contributors to An Island Polity (Renfrew and Wagstaff 1982) - namely, that for much of the second millennium the urban centre of Phylakopi was the only significant centre of habitation on Milos - excited both interest and scepticism. Some critics focused on the possibility that the form of survey was inappropriate, or perhaps simply unlucky in detecting other settlements of MC and LC date (e.g. Snodgrass 1982; Bintliff and Snodgrass 1985, 129-130; cf. response in Cherry 1983, 401-403); others, while not necessarily rejecting the apparent pattern of settlement on Milos, doubted whether they could be of wider application (e.g. Catling 1984). The fact remains, however, that neither casual discoveries nor subsequent research on the island during the past decade or more (e.g. Wiedenbein 1988) has produced any new data to cast doubt on the conclusion that Milos in late MC and early LC times was dominated by a strong primate centre at which virtually all the population of the island lived.

It was partly to provide comparative material as a check on this conclusion, which some scholars clearly found inherently implausible, that the authors, with Dr Eleni Mantzourani of Athens University, investigated in exhaustive detail an area of c. 20 square km around the site of Ayia Irini on Kea during 1983-84. The area was chosen so as to encompass the notional hinterland of that settlement, this being taken as a zone including within its bounds sufficient arable land to support the likely population resident at Ayia Irini in LC I - II when the settlement reached its maximum extent (c. 1 ha.) - an estimated 780-1250 individuals (roughly comparable to the later Classical polis of Koressos, which also embraced much the same territory). Surface artefacts within this survey region were examined by teams of fieldwalkers, who counted the numbers of finds of all periods that they observed as they walked through each and every field; these walkers were separated one from the other by approximately 15 m. Concentrations of artefacts were defined as 'sites', and spatially controlled surface collections there have allowed objective estimates of the size of each site or findspot at different periods in the past, as well as a clear picture of the range of dates to which artefacts on the surface should be assigned. At the same time, detailed observations about the locational characteristics of sites offer some insight into the reality of distributional patterns: it is obviously important to know whether the apparent absence of finds in certain areas reflects past human activities, or is a function of (for instance) dense vegetation which obscures the surface, or the presence of deep alluvial soil deposits of post-Bronze Age date. There follows a brief summary of results, solely as they pertain to the LC I period; full details are included in the final publication (Cherry et al. in press).

Aside from the results of excavations at Ayia Irini, surprisingly little was known about Bronze Age Kea before systematic survey work began on the island in the 1980s; this was especially true of the MC - LC II periods, the very time when finds from Ayia Irini suggested that relationships between Kea and Crete had been closest. Only the site of Troullos, some 500 m northwest of Ayia Irini, had produced artefacts of this date, and these seemed primarily religious rather than domestic in character (Caskey 1971, 292-295). In contrast, the survey identified finds of MC - LC II date at more than ten additional locations, none of which suggest anything other than secular use (Fig. 1). Nearly all of the prehistoric pottery found there comprised orange, red or reddish-brown fabrics of the sorts typical at Bronze Age Ayia Irini, almost certainly made locally in the northern Cyclades and probably on Kea itself. The virtual absence of obviously non-local wares is particularly striking, since imports from Crete, from the mainland and from other Cycladic islands are extremely common at Ayia Irini (a fact that was apparent from surface finds there even before excavations commenced: Cherry et al. in press, Ch. 9, n. 2); fine and decorated wares of all varieties are absent from the surface assemblages. Almost all the sherds positively or probably assignable to the MC - LC II period belong, moreover, to a very restricted range of shapes: tripod legs from plain cooking vessels, conical cups, and, above all, pithoi and large jars with relief bands of overlapping disks (Fig. 2).

Yet what is significant, for the purpose of the present discussion, is that none of these locations represents a substantial concentration of MC - LC II artefacts which obviously resulted from permanent (or perhaps even seasonal) occupation: most, in fact, have produced no more than one or two handfuls of prehistoric pottery spread over quite limited areas. Despite the use of intensive survey procedures and the expertise of several scholars to whom prehistoric Kean pottery is intimately familiar, there has not come to light in northern Kea any site comparable in size to Ayia Irini itself, and it may even be doubted that there existed any permanent settlements subsidiary to it. These paltry finds, together with their limited distribution and the restricted range of ceramic fabrics, motifs and shapes, certainly suggest the dominance of Ayia Irini; so, too, does the virtual absence of finds at some distance from the settlement, particularly in areas south of the modern port of Koressia. Other recent surface investigations on Kea, while using quite different procedures, have thus far produced results consistent with our own (Georgiou and Faraklas 1985; Galane et al. 1987).

In short, prehistoric finds are not common anywhere on Kea and, although they are more frequent in the hinterland of Ayia Irini than elsewhere, even there they occur only as low-density distributions; Ayia Irini was apparently the principal - perhaps the only - MC - LC II settlement in the northern part of the island. The considerable amount of fieldwork conducted elsewhere on Kea in the past few years has served not to weaken, but rather to reinforce, the suggestion that Ayia Irini was the only settlement of any consequence on the island as a whole.

Such pictures of Milos and Kea stand in sharp contrast to the relative wealth of information that exists for pre-eruption settlement on Thera. That this should be so is, of course, somewhat ironic, considering the impressively thick volcanic deposits that obscure the Late Bronze Age land surface in most parts of the island, and which make traditional forms of survey quite impossible. It is worth summarizing briefly the scattered evidence for MC - LC II finds outside Akrotiri itself (J.W. Sperling's excellent review of the results of 19th century excavations includes full bibliographic references, which need not be repeated here):  

Alaphouzos Quarry (Sperling 1973, 39, 56-61; Barber 1981, no. 48).

     South coast of Therasia. Fouqué (1879, 96-103) gives the impression that pre-destruction remains were commonly reached in quarrying operations in this part of the island. At least six free-standing buildings stood in an area stretching for some 125 m. The main excavated structure was at least 10 X 12 m, with six rooms on its ground floor; in one of them, a stone base supported a wooden column, while exterior corners of the house were built of ashlar blocks (on one of them what may have been a mason's mark of Minoan type). Many pithoi were recovered, some decorated. Among the finds were a press (probably for olives), barley stored both in jars and in heaps with other grains, and the bones of sheep or goats. West of the house stood a stone pillar of ashlar blocks. In general, the pottery was of styles typical of Akrotiri and neighbouring sites (i.e. LC I).

Archangelos (Sperling 1973, 13; Doumas 1983, 27).

     Hill south-west of the village of Akrotiri. Walls and pottery of the sort associated with the eruption were excavated north of the hill by Gorceix and Mamet in 1870, and Doumas has now reported LC I pottery on the hill itself.

Ayios Nikolaos (S. Marinatos 1968, 57).

     On the coast, just south-west of the site of Akrotiri; reported findspot of a stone column base.

Balos (Sperling 1973, 13-14, 53-56; Barber 1981, no. 41; Doumas 1983, 45).

    North of the village of Akrotiri, at the edge of the caldera, and c. 1 km from the site of Akrotiri. Two buildings were excavated in 1870, one a substantial house c. 10 m long and at least 9 m wide, and there may have been additional structures nearby. Inner walls of the house were plastered, and large jars or pithoi with barley, lentils, peas, and chopped straw stood in its corners; a goat skeleton lay in one room, while in a smaller adjacent structure there were many bones of goats and sheep. Fine wares of the sort well known from Akrotiri indicate occupation in LC I (Renaudin 1922).

Exomiti (Scholes 1956, 13; Barber 1981, no. 43).

     At the extreme south end of Thera. Reputed findspot of an LC I vase purchased in 1879.

Cape Koloumvos (Scholes 1956, 13; Barber 1981, no. 42).

     At the northern end of Thera. Reports of LC I vases from tombs.

Kamara or Potamos (Sperling 1973, 22-23; Barber 1981, no. 45; Doumas 1983, 45).

     About 600 m east of the site of Akrotiri. Storerooms with pithoi and other pottery characteristic of the eruption deposits elsewhere (LC I).

Karageorgis Quarries (Marthari 1982, 96; 1987, esp. 268-269; Doumas 1983, 28, Pl. 34).

     West of the village of Mesaria in central Thera, at the edge of the caldera. Graves of MC date.

Katsades (S. Marinatos 1968, 4, 59; Barber 1981, no. 49).

     West-north-west of the village of Akrotiri. Fragments of 'nicely painted store-jars', presumably LC I.

Kokkino Vouno or Mavro Rachidi (S. Marinatos 1969, 35-36; Barber 1981, no. 50; Doumas 1983, 55-56).

     Doumas has reported Late Bronze Age remains on the crest of the ridge south-west of the site of Akrotiri, with which they perhaps belong. Marinatos observed 'extensive walls', 'fragments of beautiful pottery', as well as pieces of a stone vase, metals and painted plaster.

Mavromatis Quarry (Doumas 1983, 45).

     Remains of LC I date east-north-east of the village of Akrotiri, near the edge of the caldera and c. 1 km from the excavations at Akrotiri.

Mesavouno (Scholes 1956, 13; Barber 1981, no. 46).

     The site of Classical Thera. Supposed findspot of an LC I pithos from a tomb in the vicinity of the Classical site and cemetery.

Oia Quarry (Doumas 1983, 10, 129).

     In northern Thera, near the coast. A north-facing LC I settlement has been reported.

Phtellos (Doumas 1973; Barber 1981, no. 44; Marthari 1982; 1987, esp. 368-369).

     In central Thera, near the edge of the caldera. Two LC I rooms were excavated by Doumas, and structures of the earlier part of the MC period were explored some 150 m away by Marthari; the spatial limits of occupation could not be determined in either case.

Profitis Ilias, north of (Sperling 1973, 34).

     Walls associated with pottery of types characteristic of eruption deposits elsewhere (LC I).

Although much of this evidence is still published only sketchily, it nonetheless points unequivocally to a pattern of settlement altogether different from that on Kea or Milos, and probably on other Cycladic islands too. It can hardly be fortuitous, notwithstanding the nearly ubiquitous overburden of volcanic deposits (which both preserves sites and hides them), that far more substantial remains of LC I structures and artefacts contemporary with the final destruction deposits at Akrotiri have been recognized on Thera than elsewhere. In this respect the island must be very unusual, and perhaps unique, since even making allowance for the dearth of organized surveys in the Cyclades '...one might have expected at least some evidence from elsewhere to have come to light, had it existed' (Barber 1987, 164). Furthermore, the quality and range of finds from rural Theran sites seem quite comparable to those from Akrotiri itself; they are certainly not impoverished like those from northern Kea. Scholes (1956, 26) observed some years ago that most of the material then stored in the Thera Museum came from the 'sites of Kamara, Akrotiri, and Therasia and forms a fairly homogeneous collection'. Fouqué (1879, 106) long before had speculated that vases from Therasia and Akrotiri, which were so similar that it was impossible for him to recognize any difference between them, might in fact be the work of the same potter. It was this very homogeneity of style, indeed, that permitted the 19th century excavators to assign remains to a uniform destruction horizon, at a time when the chronology of Cycladic pottery was otherwise still poorly understood.

In terms of the size and complexity of its principal settlement, Thera is likewise exceptional. To judge from present evidence, there was no other site on the island to rival Akrotiri, although it is also obvious that considerably more fieldwork is needed to arrive at a reliable measure of the size of Akrotiri itself in LC I. We might note, however, that the abundance of finds in the immediate vicinity of Akrotiri has encouraged Doumas (1983, 45; cf. Palyvou 1986, 181, n. 10) to estimate the extent of settlement as c. 20 ha. If that figure is even roughly correct, it would make Akrotiri a site that was perhaps 10 to 20 times larger than either Ayia Irini or Phylakopi, two thirds the size of Palaikastro in east Crete (MacGillivray et al. 1984, 135), and nearly one third as large as the extent of 'intensive settlement' at Knossos at this time (Warren 1984, 40).

Architectural remains at other Theran sites do not generally appear to have been built closely together, and so far there is no evidence for buildings as large and complicated as some of those at Akrotiri. The complex of structures at Alaphouzos Quarry may well have belonged to a small hamlet or village; so too might the buildings at Balos (Sperling 1973, 48); while surface remains at Kokkino Vouno hint at the existence of a 'villa of some elegance' (ibid., 48). The use of ashlar masonry at Alaphouzos Quarry points to the diffusion of Minoan architectural styles beyond Akrotiri, and the central structure is hardly of mean construction, even if the finds may 'lack the sophistication of the metropolis' (Barber 1987, 66). In several instances, in fact, the admittedly fragmentary and incomplete data allow the possibility that such sites represent something rather more substantial than merely 'minor farmsteads' (ibid., 65-66) - in which case the dispersal of population throughout the Theran landscape in LC I would seem to have been of some consequence. (The fact that agricultural features occur at some of these sites, as at Akrotiri itself, should occasion no surprise in a society whose economy largely depended on cultivation and herding, and it is not necessarily a helpful criterion in assessing the status of a site in the overall configuration of settlement. )

In summary, then, this comparative review of settlement evidence from Milos, Kea and Thera in the mid-second millennium suggests several significant points of contrast:

1. Thera seems to have been dominated by a settlement far larger and more complex than its counterparts on other islands.
2. The absence of rural sites on Milos, now strongly supported by fresh work on Kea and not contradicted by the limited data from other Cycladic islands, is not mirrored on Thera, where - despite the extreme difficulties of exploring the LC I landscape - there is growing evidence for settlement in many other parts of the island.
3. At least some of the Theran sites share elements of architectural and artefactual sophistication with Akrotiri itself, and they quite possibly represent units of settlement above the minimal level of the rural farmstead; in both respects Thera appears to be exceptional.

The interpretation of Cycladic settlement patterns has in recent years become more overtly theoretical in character (e.g. Renfrew and Wagstaff 1982; Cherry et al. in press), demonstrating in the process the utility of a long-term comparative historical framework for discussion. Clearly, there have been times in the past when population was scattered throughout the landscape rather than concentrated at large centres. On Kea, for example, it is demonstrable, in both Classical and early modern times, that political systems facilitated the acquisition of property by small landholders whose ability to participate in government was important to society as a whole. Yet at times when access to land has been controlled by a limited éite, centralized residence has been the more usual pattern. On Kea again, the concentration of settlement in a large nucleated centre was the rule throughout the later Hellenistic period; the monopolization of land-holding by oligarchies at such times must have made it difficult for small farmers to acquire contiguous parcels of land, one important prerequisite for rural residence (Halstead 1987). On the other hand, it was generally in the interests of the élite who did hold adequate amounts of land to live centrally, since this allowed them to maximize their role in decision-making for the community as a whole. But such cases differ in important ways from that of Bronze Age Thera. Settlement outside the central places of Kea in historical times has normally taken the form of single-family farms, whereas in LC I Thera it is arguable that some, at least, of the population was dispersed in larger units, such as villages or hamlets. Is it plausible to imagine that more free access to land encouraged farmers to live rurally in Thera, when there is every reason to believe that the period of the New Palaces on Crete was a time of restricted ownership of property and land under the tight control of centralized bureaucracies? The comparison with Crete is legitimate, inasmuch as administrative documents have been recognized at both Ayia Irini and Phylakopi, and at each site there is some evidence for 'special' elaborate buildings that appear to have played a central role in the affairs of their respective settlements; pot marks at Akrotiri suggest that writing was familiar there also (Palaima 1982), and more than one authority has confidently predicted the discovery there of an administrative archive on the Minoan model, together with a 'mini-palace or the residence of a local ruler' (Renfrew 1978, 418-420; Hood 1984, 34).

LM I Crete is, of course, the most obvious parallel for a Theran pattern of dispersed settlement by individual farmsteads, hamlets, villages, and perhaps even 'country-houses' (cf. Hood 1983). The striking expansion of the Minoan settlement system in Neopalatial times almost certainly reflects the reorganization of production systems under the control of the palaces (Moody 1987), and the strongly hierarchical social systems of the Bronze Age are unlikely to have promoted (or even permitted?) rural settlement by individual farmers acting on their own initiative. So does the apparent expansion of the settlement system in LC I reflect the operation of similar processes on Thera, as on Crete? It would imply that Akrotiri itself was headquarters for a centralized bureaucracy of Minoan palatial type, which in our opinion remains a good possibility even though evidence for a palace or its administrative paraphernalia is still lacking. The growth or imposition of a hierarchy of settlement on the island would be a predictable consequence of the establishment of a more hierarchically organized administration (cf. Renfrew 1982, 281). It is thus a matter of some importance for future research to provide much more information about the remains of other Theran settlements outside Akrotiri; for if dependent relationships between them and Akrotiri could be established on purely archaeological grounds, it would be possible to argue convincingly that Akrotiri operated as a central place for the island and as its seat of economic and political power, even in the absence of explicit written testimony to this effect.

It remains to be seen whether the contrasts we have drawn between LC I settlement patterns on these three islands are upheld, especially when better information becomes available from the central and eastern Cyclades, and from other islands that seem on present evidence less obviously Minoanized. If they are, then some difficult but important questions will arise. For much of their history, the political economies of the Aegean islands have been organized as primate states supporting little real hierarchy of settlement (Wagstaff and Cherry 1982; Renfrew 1982, 281); this was apparently the case on Bronze Age Kea and Milos, where a very large proportion of the populations that are likely to have been served by Ayia Irini and Phylakopi resided at these centres. So why might Thera have been different? If it is correct to think that different settlement systems reflect fundamentally different types of organization of agricultural production, why did Milos and Kea not follow the pattern of Thera and Crete? To what extent might differing histories of settlement in the MC period have influenced the organization of settlement in LC I? As yet, however, we know scarcely anything about MC Thera (Marthari 1987). Such questions tend to be stifled by the current tendency to generalize about, rather than explore the variability of, Cycladic settlement patterns, and by the assumption that observed differences are merely a product of the vagaries of the survival of the evidence or of the present state of research.

THE LC I ATRATIGRAPHIC SEQUENCES OF AYIA IRINI, PHYLAKOPI AND AKROTIRI

The effect of the eruption of the Santorini volcano on adjacent areas of the Aegean has been a hotly debated topic, but less attention has perhaps been focused on a more general question: is it feasible to attempt to relate destructions from sites in different parts of the Aegean to a single 'historical' event? Most scholars now seem to agree that the destruction deposits at Akrotiri cannot be contemporary with LM IB destructions on Crete (for a summary see, e.g., Barber 1987, 218-223; Morgan 1988, 176-177, n. 30); more recently it has been argued that LM IA, rather than LM IB, destructions on Crete might be associated with that catastrophic episode (Hood 1978; Betancourt 1978). The documentation of chronological synchronisms of this sort among destruction horizons at geographically disparate locations in Crete and the Cyclades continues to occupy centre stage in Aegean prehistory.

Most scholars, of course, are well aware of the considerable problems involved in demonstrating absolute contemporaneity of prehistoric events through comparative stratigraphic methods, and arguments are usually constructed so as to entail no such assumptions. Barber (1987, 159-160, 198), for instance, has cautiously hinted that extensive rebuilding in LC I, following roughly contemporary destructions in Ayia Irini Period V, at the end of the Second City at Phylakopi, and the early LC I Akrotiri, might reflect the use of force by Cretans in establishing their ascendancy at these island centres. His argument does not require exact synchronisms, since a Minoan conquest of the southern Aegean can be envisaged as an ongoing enterprise undertaken in stages that allow the archaeologist some chronological flexibility. The use of a specific natural event (like the Theran eruption) to explain destruction horizons is, however, a horse of a different colour: a single earthquake or eruption must occur at a single point in time. If calamities at a series of sites are to be attributed to the same natural cause, it has to be demonstrated convincingly that the levels in question are likely to be of precisely the same date. Professor Anthony Snodgrass, in his recent Sather Lectures (1987, 36-66), has discussed at some length 'the areas of incompleteness, ambiguity, and complexity in archaeological evidence' (ibid., 43) that are involved when the attempt is made to associate even a well-documented and well-dated event in Greek or Roman history (such as the sacking of a town) with the results of extensive excavations at the site in question. Those difficulties are obviously multiplied many times when such problems are set in the Bronze Age, as Snodgrass's own examples illustrate very clearly (ibid., 43-47).

At Akrotiri the sequence of LC I deposits was obviously curtailed abruptly at the time of the final abandonment of the site. Since by anyone's reckoning the LM IA period must have lasted on the order of a century, it would be of considerable help if the date of the finds on the floors of the settlement at the time of its final destruction could be fixed with more precision: was Akrotiri abandoned near the beginning, middle, or end of LM IA? The stratigraphy of Akrotiri itself provides a partial answer, since recent investigations (Marthari 1984; Palyvou 1984) show clearly that the LM IA style had already reached Thera before the town was constructed in its final form. Its final abandonment, therefore, can hardly have occurred early in the period. But how long a period of time is likely to have passed between the destruction and the beginning of LM IB? For obvious reasons, to answer this question we must turn to other sites, such as Ayia Irini and Phylakopi.

By Aegean standards, Ayia Irini offers an excellent stratigraphical sequence, exceptional in that it covers the entirety of LC I and LC II without any obvious interruptions, and that the LM IA and LM IB periods can be clearly distinguished one from another - a hard trick even on Crete. In 1972, Caskey published for the first time a synopsis of pottery of the LC I period from Ayia Irini; his Group G is an artificial assembly of well-preserved examples from several different findspots, chosen to illustrate the diversity of finds characteristic of levels contemporary with the LM IA and LH I periods. Subsequently, the phase in the history of the site to which such pottery belonged was defined as Period VI, equivalent to LM IA on Crete (Cummer and Schofield 1984, 2). The beginning of Period VI - like the LC I period, with which it is co-terminous - is marked by the first appearance of the LM IA style, its end by the first appearance of the LM IB style. As such, it includes not only the LH I period on the Greek mainland, but also final MH stages contemporary with LM IA.

The recognition of stylistic development within Period VI, as a result of work since 1972, has permitted the definition of early and late sub-phases. For instance, there are ceramic groups that include LM IA Cretan imports, but no examples of the LH I style: these are broadly similar to the earlier destruction deposits from Akrotiri that Marthari (1984; 1987, 368-370) has recently described. One of the largest such deposits from Ayia Irini comes from the middle levels of Room L.21 (Davis 1986, 62). It includes three nearly complete pots: a Cretan lentoid jug with dark-on-light decoration (Fig. 3); a Cretan teacup with dark-on-light spirals and a white band overpainted on the rim band (Fig. 4, a); and a local panelled cup with matt-black spirals on a heavy yellow slip (Fig. 4, b). Among the sherd material are the rim from a locally made Keftiu cup (Fig. 4, c); the rim of a bridge-spouted jar, imported from Milos (Fig. 4, d); and much of the rim and bowl from an imported goblet of Grey Minyan (Fig. 4, e).

The large and deep destruction deposit from House A, Room 18 (Cummer and Schofield 1984, 25, 82-87) is, however, characteristic of the overwhelming majority of Period VI deposits from the site, and it provides 'the clearest evidence of a destruction' in that phase (ibid., 32). In it are examples of both the LM IA and LH I style, as well as a range of matt-painted pottery typical of the Shaft Grave period on the Greek mainland: e.g., Polychrome Matt-painted: 'Mainland' (ibid., Pl. 64, b, e, nos. 835, 836); fine matt-painted (ibid., Pl. 64, d, no. 837); and light-on-lustrous-dark (ibid., Pl. 64, f, g). Groups of pottery of a similar character are in fact found in virtually all parts of the site, both within the town and in the substantial layers of debris that accumulated against the fortifications of the town, on both the west (Caskey 1972, 391) and east. In the latter part of the site, well-stratified layers from above Graves 23, 24, and 25, against both the east and north faces of Tower e (Davis 1986, 13) contain a broad spectrum of characteristic types: local goblets (Fig. 6, a); mainland matt-painted hydriae and jars, both monochrome and polychrome (Fig. 5; 6, b); 'Aiginetan' kraters; fine mainland matt-painted kantharoi (Fig. 6, c); as well as LH I and LM IA vessels (Fig. 6, d-f).

Others have already pointed out the broad stylistic similarities between such ceramic groups from Ayia Irini and the final destruction deposits from Akrotiri. Indeed, it was probably inevitable that the stratigraphy of Ayia Irini would be brought to bear on the question of the date of the final destruction of Akrotiri. Schofield (1984, 182) has, in fact, dated the catastrophe on Thera 'within, and probably very near the end of, Kea VI', while Marthari (1984, 132) remarked that the destruction of House A represented by the deposits in Room 18 'must have taken place when Akrotiri had already entered its late phase'. Deposits such as those in Room 18 of House A certainly do share much in common with the Theran groups, notably the presence of both the LM IA and LH I styles and the full suite of mainland 'Middle Helladic' matt-painted types, there, as on Kea, clearly of Late Bronze Age date (Marthari 1980).

This division of the Ayia Irini Period VI sequence into an earlier and later phase clearly supports an assignment of the final Theran deposits to a later stage in the LM IA period. Moreover, it has not been possible at Ayia Irini to isolate pottery groups that are demonstrably later in date than the deposits of Room 18, yet which still predate the appearance of the LM IB style. Period VIIa deposits (the earliest phase of the period) described by Schofield (1984) already contain LM IB / LH IIA imports: 'spots or dots, whether in rows as the main ornament, or as the background to another motif, such as the ivy leaf', trefoil rock-work, and double axes. Does the failure to recognize groups that are chronologically intermediate between the deposits of Room 18 and those of Period VIIa provide evidence then that Akrotiri was abandoned at the very end of LM IA?

We think not. Comparison of the deposits from Ayia Irini and Akrotiri can carry us only so far, and for the next step it is useful to turn to the LC I stratigraphical sequence of Phylakopi on Milos. The general character of the deposits there has been briefly described by us previously (Davis and Cherry 1984), but their importance warrants a brief summary. In the excavations carried out in 1974-77 by Professor C. Renfrew for the British School at Athens, strata contemporary with LM IA were encountered in nine separate trenches in different parts of the site. Careful quantitative studies of more than 30,000 sherds from these levels allowed the individual stratigraphical sequences to be placed in relative chronological order (i.e. seriated), through the application of multivariate statistical techniques. As a consequence, it was also possible to suggest the relative position within the LM IA period of specific events in the history of the site at about this time. For instance, material from the earliest floors in the 'mansion', probably the administrative centre for the LC I town, lay at the early end of the sequence, suggesting that it had been built already at, or not long after, the beginning of LM IA.

Of special interest here, however, is the fact that tephra from the Theran eruption has been recognized in levels at Phylakopi (Renfrew 1978, 406, Table III). The earliest levels with tephra lie towards the end of the sequence of LM IA strata, but not at the very end, since there do clearly seem to have been LC I strata and floors deposited later than the eruption, yet still before the earliest pottery of LM IB or LH IIA reached Milos (Davis and Cherry 1984, 156, Fig. 4). [It is relevant to report here that attempts to recognize tephra from the Minoan eruption of Santorini have proved futile at Ayia Irini. Under authority of a permit issued in 1980, ten soil samples were collected from levels still exposed in section at the site and ranging in date from Early through Late Bronze Age; these were exported to Indiana University with permission of the Greek Archaeological Service. Analyses yielded no definite traces of Santorini Bronze Age ash (c. Vitaliano per litt. August 30, 1982)].

The stratigraphical sequences of Phylakopi, Thera, and Kea are thus in general agreement in placing the destruction of Akrotiri towards the end of LM IA. The stratigraphy of Phylakopi, however, offers an unusual opportunity of examining the fine detail of chronological distinctions within LM IA. At both Akrotiri and Ayia Irini, relative chronologies rest on stylistic distinctions between destruction deposits, which by definition are composed of artefacts brought together and 'frozen' in a moment as the result of disasters; we would not expect the relatively infrequent and fortuitous occurrences of such catastrophes to have left in the archaeological record of any site a continuous record of ceramic changes. The Phylakopi sequences, in contrast, do not depend on destruction deposits, but on smaller and less impressive groups of sherds, ground continuously into the floors of the settlement or discarded regularly in its rubbish heaps; what they lack in aesthetic appeal is more than repaid by the developmental detail which their quantitative study has allowed.

The destruction levels of Akrotiri and Ayia Irini may represent 'snapshots' taken at approximately the same moment in the past, but it would be foolhardy to assume that they should be attributed to the same natural event, or associated with destructions on Crete. As yet, no deposits from either site have been published in their entirety; in the case of Ayia Irini, it will not be possible to give a fully quantified characterization of LC I deposits, since it was standard policy there to discard much of the pottery from all levels after preliminary inspection at the time of excavation. Yet it is just this kind of detailed information that is required in order to assess the likelihood that any two destruction horizons at different sites were in fact 'precisely contemporary'. Changes within LC I at Phylakopi are subtle, and it is often impossible to establish the relative chronological positions of two ceramic groups on the basis of such gross characteristics as the presence or absence of distinctive 'type-fossils': this requires the suitable quantification of the proportions of decorative motifs, fabrics, and shapes.

The preceding observations lead to several inescapable conclusions important for considerations of the date of the Thera eruption and its effects on settlements elsewhere in the Aegean:

1. The LC I sequences established at Ayia Irini, Phylakopi, and Akrotiri concur in setting the date for the abandonment of Akrotiri well after the beginning of LM IA.
2. The sequence from Phylakopi seems to suggest that the actual volcanic eruption also took place before the end of LM IA.
3. Thus the position that Akrotiri was abandoned in LM IA, but that the volcano erupted at a time when LM IB had come into use, could only be maintained by arguing that LM IB styles reached Phylakopi many years after they became current in Crete - a view which seems implausible, in the light of the plentiful evidence for regular intercourse throughout the Aegean in this period and the substantial quantities of LM IB / LH IIA pottery at both Phylakopi and Ayia Irini.
4. These conclusions shed no light on the current rift between high and low chronologies on offer for LM IA (Aitken et al. 1988; Betancourt 1987; Warren 1987): there exist no independent means of estimating the length of time in calendar years represented by the LC IA stratigraphic record at any Cycladic site. But our findings tend to support Betancourt in his belief that the abandonment of Akrotiri occurred late in LM IA, while presenting difficulties for the idea that the volcanic eruption which buried Akrotiri occurred as late as LM IB (Warren, in Aitken et al. 1988).
5. Finally, the considerable information that it has been possible to extract from subtle quantitative variation within the Phylakopi LC I sequence should alert us to the dangers in presuming that deposits which appear stylistically similar are in fact contemporaneous. Unless we pay some attention to such stylistic micro-variation both within and between individual sites, attempts to associate Cretan or Cycladic LM IA / LC I destructions with the eruption of the Theran volcano will remain thoroughly speculative.

UNIFORMITARIAN ASSUMPTIONS AND 'SUPPOSITIOUS CORRELATIONS'

The preceding case studies were chosen to illustrate the dangers of uniformitarian assumptions in the interpretation of archaeological data. While it is obviously the case that LC I settlements on Thera, Kea and Milos in LC I were in many respects very similar to each other (particularly in their imitation of Minoan styles and in the adoption of Cretan technology), the patterns of settlement on these islands seem to have been far from uniform. The differences we have described cannot be dismissed as accidental, partly because the (largely negative) Kean and Milian evidence derives, not from haphazardly accumulated finds, but from systematic and intensive exploration, and partly because the Theran evidence comes from a geological setting that in general militates against site discovery. The inferences to be drawn from such contrasts may, of course, be found unconvincing, and it should be clear that our suggestions have the status merely of working hypotheses, to be considered as better data also begin to appear from other Cycladic islands. Similarly, we have also argued that the detailed examination of ceramic deposits can establish long and useful stratigraphical sequences and detect stylistic micro-variation that is not necessarily apparent at first glance. This leads us to the opinion that it is a mistake to draw far-reaching conclusions about the exact contemporaneity of destruction levels in the Aegean from purely stylistic evaluations that are based on anything less than the fully quantified description of all the available finds from each site.

Some, perhaps most, Aegean prehistorians seem to feel that their proper role is attempt 'historical reconstruction' - that is, to derive historical narrative from purely archaeological observation in a period which, by any sensible definition of the term, is in fact fully prehistoric. Yet, as Snodgrass wrote recently, 'It is surely clear that any kind of "historical" narrative, for a culture in which any of the dates may be even fifty years out, let alone two hundred, is an impossibility. The very language of political developments and military episodes, in which narratives of the Aegean Bronze Age have for long been couched, seems inappropriate.' (1985, 37.) It was this same inherent chronological imprecision which led Renfrew (1979, 582) to complain of the process of fallacious reasoning which he dubbed the 'method of suppositious correlation': 'Disparate events simply should not be cross-correlated unless their simultaneity can independently be documented. Indeed there is the risk that the volcanic eruption may on occasion replace the migratory horde as an easy explanation to be assumed on convenient occasions by those who will not critically evaluate their evidence.'

If that is so, then Cycladic and Minoan archaeologists who hope to achieve a form of narrative history for the Bronze Age would be well advised to devote more effort to the development of fine-grained chronologies which are truly capable of tracing associations between archaeological events that cannot be directly linked through stratigraphy. To proceed with business as usual, pretending that the simultaneity of past events can be established with greater precision than the data in fact allow, is intellectually dishonest. Our own feeling is that the relatively coarse chronologies available at present are suited more to the study of broad patterns and processes of human activity (e.g. settlement patterns, which are the result of repeated behaviour), than to the documentation of specific events. Clearly, both types of information would be desirable; but we need to do a better job of disentangling the historically specific from the general.

Addendum

In the light of the discussion following our paper, two points evidently require further clarification. First, we used the term 'settlement' to refer generally to any form of residence on the land, not only to aggregations of population in towns, villages or hamlets. Even in this general sense, residential settlements outside the centres of Ayia Irini on Kea and Phylakopi on Milos seem, quite simply, not to exist in the early LBA. The Kean distribution, it should be emphasized, does not, we think, represent loci of human occupation, but merely findspots that have yielded very modest quantities of MC - LC II material. On Milos and Kea, where the Bronze Age land surface is not obscured from view, rural sites have not come to light in the course of very thorough surface survey on both islands. That so many are already known on Thera, even though only a tiny fraction of the pre-eruption surface has so far been exposed through quarrying, erosion or other disturbances (cf. Aston and Hardy 1990), serves merely to reinforce the contrast we wish to make: if the patterns of settlement on these three islands were at all comparable, we would expect to have found either far more sites on Milos and Kea, or far fewer on Thera. It therefore seems to us relatively uncontroversial, on the basis of current evidence, to assert that Milos and Kea were characterized by a strongly primate pattern, with little or no residence in the hinterlands of their Bronze Age centres, while Thera reveals a more complex picture of dense and diffuse rural settlement, organized in villages, hamlets, villas and isolated farms. What lies behind this contrast, however, is a more complex matter. The Theran pattern is certainly reminiscent of contemporary settlement in Neopalatial Crete, but is it not dangerous to assume, as does Cadogan (1990), that it therefore reflects the extension of Minoan peace and security or even 'Cretans coming to settle'? We surely need a more sophisticated analysis of the relationship between centralized and dispersed residence and economic and political circumstances.

Second, we want to reiterate our belief that the eruption occurred at a mature stage of LM IA, but not at its very end (as some delegates to the Congress seem to have assumed). Our evidence stems from quantitative studies of LC I levels at Phylakopi, where floors and strata were laid down after the first tephra-bearing levels, but before the first with LM IB or LH IIA material. It should be noted that, while it is certainly possible to isolate certain motifs on local Milian painted pottery which are restricted in their occurrence either to an early or to a late stage in the LC I sequence (Davis and Cherry 1984, Fig. 5), the post-eruption LM IA levels are distinguishable mainly in terms of the relative proportion of particular ceramic categories that are also found earlier, rather than of distinctive 'type-fossils'. In the light of this stratigraphic and ceramic evidence, and considering the intensity and regularity of intercourse throughout the Aegean in the early LBA, we find it very hard to accept the idea that the later LC I horizons in the islands might in fact be subsequent to the passing of LM IA on Crete.

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

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