The Eruption of Thera: Date and Implications

At the same time, however, several leading archaeologists have published critical responses to all such proposals, and defended the conventional chronology, which dated the eruption c. 1500 BC. The most generous of the responses, by Cadogan, is prepared to countenance a slightly earlier calendar date for the eruption, but merely by placing the geological event a little earlier in the LM IA ceramic period. In all cases the conventional interpretation of the absolute ceramic chronology, derived from the usual interpretation of correlations with Egypt, is accorded overriding precedence. The radiocarbon evidence is dismissed as yielding too large a date range to be of use, or regarded as compatible with either a high or a low chronology; while various climatic data are rejected as not necessarily associated with the eruption of Thera.

As such a controversy implies, many of the essential issues are not clear-cut. This paper attempts to stand back and assess the relevance, quality and precision of the various classes of data pertaining to the absolute date of the eruption, so as to reach the best solution on the existing evidence. This process of putting the house in order seems an important requirement before consideration is given to the new evidence to be presented at this conference - such as the Oxford AMS dates - in order that the significance of these new data does not merely become lost in the present quagmire.

From a survey of the data by class of evidence, it is concluded that:

(a)    The conventional interpretation of the ceramic correlation data is debatable, and a higher interpretation is both possible and plausible; however, overall, the archaeological correlations are not precise and cannot necessitate the conventional chronology.

(b)    Despite some technical problems, the statistical analysis of the large set of radiocarbon dates on short-lived samples from Akrotiri shows that they provide the best, most precise evidence currently available, and strongly support a later 17th century BC (and not 16th century BC) date for the later LM IA period.

(c)    A range of data reveals a northern-hemisphere-wide climatic event at a similar, later 17th century BC date; but whilst this is suggestive, there is as yet no direct evidence to link it with the eruption of Thera.

The existing balance of probabilities must be held to support a later 17th century BC date for the later LM IA eruption of Thera. In turn, the LM IA period must be placed in the 17th century BC. This conclusion has major implications for Cypriot, and hence Near Eastern chronology. Aegean contacts with Europe are also affected. A minor chronological revolution may be at hand. The confirmation of this view, or the spread of only further controversy, may well depend on the series of AMS dates from Oxford and elsewhere. From past radiocarbon dates, and the shape of the calibration curve between c. 1625 - 1525 BC, there appear two possible results: an ambiguous result compatible with both a high and a low chronology, or a result compatible with the 17th century BC radiocarbon analysis presented in this paper. We shall have to wait and see!

INTRODUCTION

For a variety of reasons, direct and indirect, a number of scientific and archaeological papers published in 1987 and 1988 proposed a 'high' 17th century BC date for the Bronze Age eruption of the Thera volcano, and hence for the beginning of the Aegean Late Bronze Age (Betancourt 1987; Hammer et al. 1987; Baillie and Munro 1988; Aitken 1988; Michael and Betancourt 1988a; Manning 1988). Several of these papers took their lead from the earlier arguments of Michael (1976), Kemp and Merrillees (1980) and La Marche and Hirschboeck (1984). At the same time, however, the literature saw several of the leading scholars in the field publish critical responses to all such propositions, and defences of the conventional chronology which dated the eruption at c. 1500 BC (e.g. Betancourt and Weinstein 1976; Warren 1984, 1985, 1987, 1988; Cadogan 1978, 1987; Hankey 1987). In these critiques, both Warren and Cadogan question the necessary relevance of the climatic data, Hankey questions the accuracy of the carbon-14 evidence, and Warren would dismiss the carbon-14 evidence as yielding too large a date range to be of use. In each case the conventional interpretation of the archaeological evidence has been given overriding precedence. Most recently, a scientist, Pyle, has highlighted the view that there is no necessary causal link between the evidence from ice cores and tree rings for a climatic event in the 17th century BC and the eruption of Thera (1989).

No one denies the importance of the Thera date to Aegean archaeology. However, as the existence of such an ongoing controversy implies, many of the essential issues are not clear cut, and are open to a variety of interpretations. Therefore, what appears of greatest utility now is not another partisan argument, nor a detailed recapitulation of all the data so that the point is lost in detail, but rather an attempt to stand back and briefly assess in a non-partisan fashion the relevance, importance, and then implications of the existing body of data for the absolute (calendar) date of the eruption of Thera. Moreover, of relevance in the present context, such a process of putting the house in order is an important requirement before consideration is given to the new evidence to be presented at this conference - such as the Oxford, Simon Fraser, Zürich, Copenhagen and Heidelberg radiocarbon dates - in order that the significance of these data do not merely become lost in the present quagmire.

RELATIVE DATE

The preliminary question is, of course, the relative date of the eruption in terms of the Aegean ceramic sequence. From the testimony of the numerous mature Late Minoan (LM) IA ceramics found buried under the eruptive deposits at Akrotiri on Thera (Marinatos 1968 - 1976; Doumas 1983, 1989; Marthari, this volume), the eruption has usually been placed close to. the end of the LM IA period (e.g. Niemeier 1980; Doumas 1983, 138 - 139; Betancourt 1985, 133). As discussed at the previous Thera Congresses, the only other possibility seems to be the existence of a significant interval between the abandonment af Akrotiri in late LM IA, and the eruption, allowing the possibility of an LM IB date for the eruption itself (e.g. Warren 1984, 492; Manning 1987, 70 - 71). This latter view might draw same support from the apparent archaeomagnetic association between the eruptive deposits an Thera and the fired remains of the LM IB destruction levels on Crete (Downey and Tarling 1984), or an interpretation of the fresco evidence (Morgan 1984).

However, the evidence from the stratified finds of Theran tephra and pumice at a number of Aegean sites supports an LM IA date for the eruption, and seems to rule out a later LM IB date (e.g. Vitaliano and Vitaliano 1974; Doumas 1983, 137 - 139; Marketou 1988; Betancourt and Davaras 1988, 218; papers in this volume by Soles and Davaras; Marketou; Betancourt et al. ; and Warren and Puchelt). Furthermore, a careful study of the occurrence of Theran tephra at Phylakopi on Milos in fact suggests that the eruption not only occurred in the LM IA period (Renfrew 1978, 414 - 415), but that it occurred at least a little before the end af the Late Cycladic I / LM IA periods (Davis and Cherry 1984, 154, 156, 158). Cadogan has also very recently begun voicing a similar opinion, that the Minoan ceramics at Akrotiri possibly do not represent the end of the LM IA period (1987; 1988, 99), although he has not as yet substantiated this view. However, against any significant 'archaicizing' of the Akrotiri destruction level LM IA assemblage, it must be noted that MM IIIB - early LM IA imports were already reaching the site before the major, penultimate, early Late Cycladic I destruction (Marthari 1984, and this volume). This suggests that the much later abandonment of the subsequently rebuilt settlement, containing quantities of advanced LM IA ceramics, must have occurred at least well into the LM IA period (see also Davis and Cherry, volume one of this Congress). Thus occasianal 'archaic' features are either heirlooms, or interesting stylistic survivals (e.g. Marinatas 1972, Pl. 60.c). Nonetheless, this issue does highlight Popham's observation that a more rigorous definition of the LM IA style is necessary at Knossos, and for Crete in general (1977, 195).

The conclusion at this time is that the abandonment of Akrotiri, and the eruption, may be dated during the later LM IA period.

ABSOLUTE DATING

An important distinction must be made between the direct dating evidence, from radiocarbon dates from the destruction level at Akrotiri or from ceramic correlations for the LM IA and surrounding periods, and the indirect evidence from datable climatic effects which only might be causally related to the eruption of Thera. It is of course frustrating that it is the indirect, or proxy, evidence which is so remarkably precise!

DIRECT DATING EVIDENCE

• Ceramic evidence:

The basic problem is the lack of any certain LM IA pottery from historically datable contexts in the Near East. Thus, as Warren (1984, 492) stated, 'Absolute dating of these ceramic periods from artefactual evidence is far less precise than we would like.' Therefore, before considering the possible or debatable evidence for the LM IA period, the first step appears to be to attempt to try and delimit the range of possibilities for the LM IA period from a consideration of the evidence for the preceding and succeeding ceramic periods. The overall aim must then be an assessment of the implications, and quality of the date.

In a number of recent contributions Warren has argued that the Khyan lid from Knossos limits the upward mobility of the LM IA period, as he holds that this lid represents a terminus post quem for the Middle Minoan (MM) III (A) period (1984, 492; 1985, 149; 1987, 206 - 207; 1988, 176). However, it is simply not possible to be so confident of the find context of this lid, as argued by Palmer, Pomerance, Betancourt, Walberg (see Manning 1988, 31 with refs.), Strøm (1980), and Raison (1988, 206 - 209, 234, 238 - 239; also. cf. Cline 1987, 31 and refs.; Michael and Betancourt 1988a, 169 - 170). For instance, in the original excavation notes, the lid is in one reference described as found 'superimposed on a deposit containing a great deal of charcoal' (Raison 1988, 208 n. 179), which, as this charcoal layer is probably to be assigned to MM IIIA (e.g. Warren 1987, 207), would leave the lid as later than MM IIIA (LM IA?), and sealed only by the LM III wall. (The integrity of the MM IIIA date is also questionable as the pottery was mixed; nor is pottery with white spots on a dark ground exclusively MM IIB / IIIA: cf. e.g. Catling et al. 1979, Fig. 18.49, 95 - 98 and refs. at pp. 25 no. V.49 and 28. V.95 - 98; Marinatos 1972, Pl. 60.c.) In contrast, elsewhere the lid is merely described as from the deposit immediately beneath the LM III wall, without indication of internal grading (Raison 1988, 208 and n. 179). It is thus inappropriate to employ the lid as a chronological keystone. Instead, at Knossos, the Egyptian alabaster alabastron fragment, probably of Second Intermediate Period attribution from an MM IIIB context (Betancourt and Weinstein 1976, 337 and n. 73), seems to provide the best terminus post quem for LM IA, as MM IIIB must at least extend into the Second Intermediate Period, c. 1801 / 1759 -1570 / 50 / 39 BC (Kitchen 1987, 50; Wente and van Siclen 1976; against the recent ultra-low Egyptian chronologies (e.g. Krauss 1978; Helck 1987), see Murnane 1983). At Akrotiri, the presence in the destruction level of probable Syro-Palestinian MB II (B - C) vessels (Warren 1979, 88 - 90, 106 - 107; Marinatos 1976, 15, 30, Pl. 19b, 49b; Niemeier, this volume) sets a compatible terminus post quem, and likewise suggests an actual date within the Second Intermediate Period (as Warren 1984, 492). Unfortunately, however, there is a lack of complementary MM III material from the Near East, with the few possible items (e.g. Schaeffer 1932, 5 - 6, Pl. 7.1) lacking tightly dated contexts (also Betancourt 1985, 112; Manning 1988, 29 - 30).

At the lower end of constraint, several pots from the succeeding LM IB or equivalent Late Helladic (LH) IIA period occur in Egyptian contexts related to the period from the early 18th Dynasty to the reign of Tuthmosis III (Betancourt and Weinstein 1976, 337 - 339; Kemp and Merrillees 1980, 253), c. 1504 / 1479 - 1450 / 25 BC (Casperson 1986; Kitchen 1987, 40), or from even later contexts (Hankey 1987, 46; 1973, 107). Such data set our initial terminus ante quem for the LM IB period, and hence the close of the LM IA period. The important point is that the length of the 'ante' is unknown; time lags are common. The real date could be much earlier. The basal position is set by the LH IIB (roughly contemporary with LM II) squat jar from Kahun in a context dated before the end of the reign of Tuthmosis III (Hankey and Tufnell 1973). Thus the preceding LM IB period must have at least ended before the late years of Tuthmosis' reign. Again, the real date could have been much earlier (Kemp and Merrillees 1980, 253 and n. 523). Moreover, Wachsmann (1981, 127 n. 5) recently refers to a view that the coffin with this jar in fact dates to the earlier part of Tuthmosis' long reign. Thus LH IIB might date from before, or at least by, c. 1500 / 1480 BC, with the LM IA and IB periods before this. (Reinterpretation of a tomb at Saqqara which contained two Mycenaean imports may also support such views: Miss J.D. Bourriau, pers. comm.) Betancourt has of course gone much further, arguing that the Aniba alabastron from an early 18th Dynasty context (Tuthmosis III at the latest: Kemp and Merrillees 1980, 243 - 244, 253 - 254; Weinstein 1983) is an imitation of LM IIIA 1 (1987, 46 - 47; Michael and Betancourt 1988a, 170, 1988b). In this case, the LM IIIA 1 period would have begun no later than the reign of Tuthmosis III, and the LM II, LM IB, and LM IA periods were all earlier, consistent with the evidence above. The Sellopoulo scarab of Amenophis III would then set the approximate close of the long LM IIIA 1 period, consonant with the majority of the inscribed objects of this ruler or his wife coming from subsequent LH IIIA 2 - IIIB contexts (Cline 1987, Table 1). It is not incompatible with a high chronology (Manning 1988, 36 - 37). On the other hand, it must be noted that Warren has disagreed with each of the above interpretations (1987, 1988).

The determination of the dispute would be simple if there was clear evidence for the earliest possible date for the start of LM IB. However, this does not exist at present. At best, the find of two Red Lustrous wheel-made spindle bottles in LM IB contexts on Crete (Hawes et al. 1908, 42, Pl. 8.25; Watrous 1985, 7 - 8) suggests that the close of this period at least overlaps with the Syro-Palestinian LB I period (cf. Manning 1988, 34; also the fragments of 'Canaanite' jars from Zakros: Kopcke 1987; 256 n. 9), or, depending on choice of Egyptian Pharoah chronology, some point in the first half of the 16th century BC. On the other hand, there is the LM IB sherd from Abydos tomb 328 (Kemp and Merrillees 1980, 232 - 236, 252 - 253). This tomb is almost certainly pre-18th Dynasty, or Hyksos. Warren (1985, 149; also Kemp and Merrillees 1980, 238) is correct to point out that one vase potentially has later parallels (to Tuthmosis III), but it is a general shape, and also has 13 - 17th Dynasty parallels (Kemp and Merrillees 1980, 238). Thus as the other evidence supports a Hyksos date for the tomb, and this vase can also be so interpreted, the presence of the LM IB sherd most reasonably provides evidence for a pre-18th Dynasty correlation for the LM IB period. These data are consistent with a date for some part of the LM IB period in the first half of the 16th century BC. (The publication of LBA Aegean sherds from the Egyptian Exploration Society excavations at Kom Rabia in Egypt by Miss J.D. Bourriau will provide new data on these problems. In particular, the Aegean sherds from a deposit (530) belonging to the very beginning of the 18th Dynasty (Ahmosis and Amenophis I) will set a terminus ante quem of c. 1570-1500 BC for these styles. While Miss Bourriau (1989) initially suggested an LH IIB attribution for this material, an LM I assignation is perhaps more likely; much depends on whether it is LM IA or LM IB ! I wish to thank Miss Bourriau for discussing this material with me.)

Within the bounds of the preceding discussion, we might now consider the evidence for LM IA chronology. The first item of note is the sherd from Kerma which several authors believe is LM I (Smith 1965, 39 - 40 and Fig. 59; Kemp and Merrillees 1980, 244; Warren 1985, 150). The find context is imprecise, but sets a terminus ante quem early in the reign of Kamose, or at the beginning of the 18th Dynasty. Thus this is evidence for a Second Intermediate Period date for some part of LM I.

The second item is the well-known, but little understood El-Lisht jar (Kemp and Merrillees 1980, 220 - 225). The Tell-el-Yahudiyeh technique decoration includes dolphins, and so a Minoan stylistic influence is usually considered present; although it must be noted that dolphins occur on other Tell-el-Yahudiyeh vessels from Dashur and Tell-el-Dab'a (J.D. Bourriau, public lecture, Cambridge 28/5/89) and the motif thus may well be considered of local Syro-Palestinian origin. However, even if we assume that there is an Aegean connection (for dolphin decorations are not exclusive to Crete; for example, several occur on pots from Akrotiri on Thera), the problem then is the choice of Minoan comparisons. In 1985 Warren was happy with the MM IIIB / LM IA or LM I jar from Pachyammos (1985, 149) employed by Betancourt (1987, 47), which, as the El-Lisht jug was found in a 12 / 13th or 13th Dynasty context (Kaplan 1980, 94; Kemp and Merrillees 1980, 224 - 225), would push the MM III / LM IA border back before c. 1650 BC. However, Warren now wishes to compare instead a slightly earlier MM III (A) pithos decoration (1987, 206). The conclusion must again be that the data are inconclusive.

It is submitted that the archaeological evidence discussed above, and the wider corpus discussed previously (Manning 1988, 26 - 38), are consistent with a higher chronology than usually accepted. However, it is also recognized that there are many problems, and in most cases a different interpretation is possible. Thus it is arguable that Betancourt is justified to question the conventional stylistic attributions and correlations, but that Warren is also right to dispute various of these positions, and so on. Such is art-history: there is no proof, only interpretation. Similarly, for the important items whose stylistic attributions are not debated, it cannot be denied that their find contexts are uncertain, or less than restrictive. Examples are the Khyan lid (above), the LM IB sherd from Tell Ta'annek for which the usual Tuthmosis III year 23 context is based on surmise and no evidence (Hankey 1987, 45), or the squat jar from Kahun (above).

The end point is that each conventional archaeological datum at issue is contestable to varying degrees (Betancourt 1987; Manning 1988). Therefore, the conclusion must be that the ceramic evidence is not precise, only suggesting various wide chronological ranges, and usually merely a terminus ante or post quem.

• The radiocarbon evidence (pre-Congress):

A number of radiocarbon dates on short-lived samples from volcanic destruction level contexts on Thera should provide a direct date range for the late LM IA period, and, approximately, for the subsequent eruption. As these dates are all representative of the same chronological horizon, and certainly so in terms of the realms of the precision technically possible, they should be considered as a set, not separately (Manning 1988, 40). The total set of 18 dates (e.g. Manning 1988, Table 4; Hammer et al. 1987, Table I, adding P-1888, P-2560, P-2561 and P-1697) has some outliers (as is to be expected; for instance on a 68% one-sigma basis, up to six dates might be expected to fall outside the average range), but has a good, consistent central set. Statistical analysis is necessary.

Aitken (1987), Hammer et al. (1987) and Michael and Betancourt (1988a) have recently presented averages and the reasons for the approximate validity of the dates. However, to varying degrees, each of these analyses is less than satisfactory and needs re-assessment. Employing a Ward and Wilson analysis (1978; with ± 14 year calibration error: cf. Pearson et al. 1986, 912), the data set comprising Aitken's average is found to have an unhappy test statistic of T' = c. 18.13, barely less than χ²_10;0.05 = 18.3. The reason is the inclusion of the aberrant low dates of P-1888 and P-2794. This average is unsatisfactory. The average of Hammer et al. is higher and statistically better, but still has an uncomfortably high test statistic of T' = 15.12 compared with χ²_13;0.05= 22.36. The explanation is the inclusion of the aberrantly low P - 2794 and high P-2792. The former is most notable, as this low date deviates by ≥ 2.5σ from their mean. Moreover, because of its small error, its aberrant inclusion significantly biases and lowers their overall weighted mean (unlike the other undersized inclusions, with large errors, which have little effect at 1/ σ²). Furthermore, P-2794 is in fact a soil sample, and not strictly short-lived. It probably ought never to have become included in the short-life date sets under discussion! Its exclusion would produce a higher average. The objection to the final average, that of Michael and Betancourt (1988a), applies also to the preceding averages. This is that the dates selected to be averaged (in this case nine) are selected on an entirely subjective basis, and not on any generally accepted statistical criteria. Thus we read: 'Two of the four remaining samples (P-2560 and P-2561) are obviously deviant, and on the basis of excellent consistency on the part of the nine selected samples, we judge P-1888 and P-2794 to be also deviant.' (Michael and Betancourt 1988a, 171.)

As described previously (Manning 1988, 38 - 46), a new analysis is called for, employing a statistical treatment. If a clustering procedure is employed (Wilson and Ward 1981) then from the 18 dates a coherent and significantly large central group of 12 dates is determined (Manning 1988, Table 5; or Hammer et al. 1987, Table 1 excluding P-2792 and P-2794 -note as deviants, for, as above, P-2794 might also be excluded on principle as not necessarily short-lived). The weighted mean of this set of 12 dates (Ward and Wilson 1978) is c. 3321 ± 17.5 BP, with T' = 4.1<χ²_11;0.05= 19.7. The calibrated date at ± 1σ on the Pearson and Stuiver(1986) curve calculated using the programme of Stuiver and Reimer (1986, rev. 2.0), is c. 1670-1662, 1647-1606, 1554-1543 BC. It must be noted that the small mid-16th century BC range is statistically unlikely as the appropriate section of the calibration curve mainly lies outside the total 1σ boundaries of the weighted average (Pearson and Stuiver 1986, 841). The probability of the 1647-1606 range is 62%, compared to only 20% for the 16th century BC range. Moreover, the 'finer' decadal calibration curve of Stuiver and Becker (1986) which is recommended as the more appropriate for short-life dating (Stuiver and Becker 1986, 863), is also decisively in favour of a 17th century BC date range, yielding (via Stuiver and Reimer 1986) c. 1686-1672, 1658-1654, 1640-1601, 1560-1536 BC, with the 1638-1602 range the most likely (47%).

Therefore the existing radiocarbon evidence provides a precise later 17th century BC date range for late LM IA, contrary to the view that the method is too imprecise to be of significance, or can accommodate both a high and a low chronology (e.g. Cadogan 1987; Warren 1987, 1988). In addition, contra Warren (1987, 208 - 210), it has been shown elsewhere that the other Aegean Middle Bronze III - Late Bronze I radiocarbon date series are compatible with this position (Manning 1988, 49 - 55). Other more tenuous radiocarbon data are also consistent with a probable 17th century date (Stanley and Sheng 1986; Zouridakis et al. 1987, 63; it must be remembered that the single radiocarbon date of 3110 ± 160 BP recently reported for the layer of peat some 30 cm above a Thera tephra layer near Sardis in Turkey (Mellink 1989, 117) merely sets a terminus ante quem for the tephra - the real date for the deposit of the tephra some 30 cm below this peat should be several centuries earlier, depending on the stratigraphy.)

In recent discussions, this leaves only two question marks remaining over the radiocarbon evidence. The first is the well-known problem of whether the samples might have been affected by emissions of carbon dioxide from the volcano (e.g. Olsson 1987, 18 - 22). This is of course possible, but unlikely. The first reason is the consistency of the large central set of dates, whereas volcanic contamination, when present, is highly variable (see e.g. Olsson 1987, 20 and refs.; the effect might thus explain the two extraordinarily early outliers: P-2560 and P-2561). The second reason is that the effect is usually extremely localized (Taylor 1987, 131 - 132; Aitken 1987, 19), and perhaps only affects plants growing in shielded locations (Michael and Betancourt 1988b, 180). Lastly, the windy nature of the modern, and presumably ancient, island of Thera would promote rapid atmospheric mixing, further making it unlikely that a large number of dates from different plants from different storage areas (cf. Betancourt 1987, 46) would all be consistently affected. This leaves the remaining problem: the lack of carbon-13 normalization for most of the Pennsylvania dates (only three in the central set of 12), or consistent NaOH pre-treatment (Olsson 1987, 28 - 30). Moreover there is also a worrying discrepancy between the few δ¹³C data reported for some dates in 1977 and 1981 by Pennsylvania (P-1697, P-2791, P-2793, P-2795), and the values found at Copenhagen (Friedrich et al. 1980). The Copenhagen values (requiring an upwards correction) would support the high 17th century BC chronology even more strongly than the Pennsylvania dates. However, other than merely increasing the error margins (Olsson 1987, 24 - 26, which, following her suggestion at p. 30 to assume a δ¹³C correction error of ± 3 for the non-normalized dates, would raise the above weighted average slightly to c. 3328 ± 21 BP, calibrated on the Pearson and Stuiver 1986 curve (via Stuiver and Reimer 1986): c. 1674-1610, 1551-1548 BC - correction to Manning 1988, 66 n. 14, with a 92% probability for the range 1675-1609 BC), we must await new data with full carbon-13 data (and alkali pre-treatment), before this problem can be overcome. The weighted average from the five dates with carbon-13 normalization in the central 12 (P-2791, P-2793, P-2795, K-3228, K-4225) is c. 3351 ± 32 BP. Calibrated on the Pearson and Stuiver (1986) curve this yields: c. 1687-1618 BC. The weighted mean of the five dates on samples pre-treated with NaOH in the central 12 (P-1889, P-1892, P-1894, K-3228, K-4225) is c. 3330 ± 25 BP. Calibrated on the Pearson and Stuiver (1986) curve this yields c. 1676-1609, 1552-1547 BC.

• Indirect, or possible, dating evidence:

The study of the long term climatic records provided by the annual layers in an ice core from Greenland, and by tree rings from two continents and several countries, all combine to show that there was a major climatic event in the later 17th century BC (La Marche and Hirschboeck 1984; Hammer et al. 1987; Baillie and Munro 1988; Hughes 1988; Baillie, this volume; see also Bernal 1987, 42 and n. 18). From the pattern of historical correspondences (La Marche and Hirschboeck 1984; Hughes 1988, Table), volcanic sulphur dioxide in the ice core (Hammer et al. 1987), and the probable causal models (Sear et al. 1987; Baillie, this volume), it is very likely (but of course not certain: Pyle 1989) that the origin was a large volcanic eruption in the northern hemisphere. The question is whether this eruption was Thera, especially as the carbon-14 data suggest a very compatible date range.

To start with, it must be observed that there is as yet no method for 'fingerprinting' such data (by elemental or isotope analysis), or for linking such events with specific eruptions. Nor do all large eruptions appear to leave traces on the ice-core or tree-ring records. Thus there is no direct link, and scepticism is justified (Warren 1984; Pyle 1989). Another suitable large volcanic eruption may be found, with no means of establishing whether it, or Thera, was the cause of the climatic data. On the other hand, contrary to the minimizing slant of Pyle (1989), the eruption of Thera does appear to have been a rather special event in the period since the last Ice Age. Pyle suggests that a number of eruptions such as Thera have occurred since AD 1800, a comparison based on an assigned Volcanic Explosivity Index of 6 for Thera (1989, 88 - 89). However, such a value is a modern 'guesstimate', as the necessary observational data for the calculation were not recorded at the time! Instead, one might consider the view of Watkins et al. (1978; also see Sparks et al. 1983/84, 165 - 167) that the eruption of Thera was one of the largest post-glacial eruptions, and that 'eruptions of this magnitude are yet to be chronicled.') The recent discovery of a significant remnant of Theran tephra in Turkey only further supports the existing arguments in favour of the enormous scale of the eruption (Sullivan 1988; see also Stanley and Sheng 1986). From such a viewpoint the link between this special, very large eruption, and the only major volcanic-climatic event recorded from the mid-second millennium BC, appears more reasonable. But it is also clear that we may never know for certain.

CONCLUSIONS AND IMPLICATIONS

From the above discussion three conclusions may be drawn as regards the date of the late LM IA period and the eruption of Thera:

1. The conventional interpretation of the archaeological evidence is debatable (above; Kemp and Merrillees 1980 versus Warren 1985, etc.), and a higher interpretation is possible (Betancourt 1987; Manning 1988); however, overall, the archaeological correlations are not precise (as often thought), and cannot necessitate the conventional (lower) chronology.
2. Despite some problems, the radiocarbon data are the best, most precise, evidence currently available, and strongly support a later 17th century BC date for the late LM IA period.
3. A range of indirect evidence appears suggestive of a similar, later 17th century BC date.

The existing balance of probabilities must be held to support a later 17th century BC date for the eruption of Thera. In turn, the LM IA period must be placed in the 17th century BC (as Betancourt 1987; Manning 1988), and a new 'high' Aegean Late Bronze Age chronology is both necessary, and plausible. Acceptance of this position has major chronological and historical ramifications, including outside the Aegean. For instance, LM IA correlations with Late Cypriot I contexts (Betancourt 1985, 130), or a Late Cypriot I White Slip I bowl found with Middle Cycladic II - Late Cycladic I / LM IA pottery in a volcanic destruction context in the Akrotiri ravine on Thera (Fouqué 1879, 125, Pl. 42.6), require a 'high' Late Cypriot chronology (i.e. no later than Merrillees 1977). This in turn may have an impact on the presently contentious area of Syro-Palestinian Middle Bronze II - Late Bronze I chronology. For instance, a high Middle Cypriot III - Late Cypriot I date is difficult to reconcile with the very low chronology of Bietak (1984), or the low dating of Alalakh Levels VI-V by Gates (1981, 1987). On the generally accepted assumption that they do represent Aegeans, the conventional interpretation of the paintings of the Keftiu in several tombs at Thebes in Egypt is another subject that is both a challenge to, and challenged by, the new chronology. For example, one may immediately note that the kilts in the Rechmire and Mencheperresonb tombs are best compared with those of the cup bearer and two other figures in the Procession Fresco from Knossos (e.g. Wachsmann 1987,44). As this fresco probably dates from the LM II - IIIA period (see: Hawke Smith 1976, 70 - 71; Hood 1978, 66; Boulotis 1987; Popham 1987, 298), it appears consistent with the correlation of these periods with the reign of Tuthmosis III. Many of the other aspects of the Keftiu paintings in fact are, on examination, less than diagnostic (e.g. Schachermeyr 1960; Hankey and Warren 1974, 146; Kemp and Merrillees 1980, 280; Wachsmann 1987, 60, 66 - 67, 72 - 73). A high Thera chronology would also affect the chronology and interpretation of Aegean contacts with Europe. A minor chronological revolution may be at hand.

The confirmation of this view, or the spread of only further controversy, may well depend on new series of AMS dates from Oxford, Copenhagen, Sinion Fraser, Zürich and Heidelberg. From past radiocarbon dates, and the shape of the calibration curve between c. 1625-1525 BC, there appear two possible results: an ambiguous result compatible with both a high or lower chronology (Aitken 1988, 169), or a result compatible with the 17th century BC radiocarbon analysis above. We shall have to wait and see!

Addendum

For a general review of chronology in the light of the new evidence presented at the Congress, see Manning 1989, 1990. Those papers argue that the new evidence favours but does not prove the 'early' chronology. Further evidence, such as the radiocarbon dates for organics associated with Thera tephra in Turkey, promised in the paper by Sullivan (this volume), is therefore keenly anticipated.

Of greater concern, however, is that apart from usual problems of inter-laboratory variations and deviants, it must be noted that three fundamental problems emerged at the Congress.

1. The non-linear properties of radiocarbon calibration and the taphonomic irregularities involved were highlighted by Weninger. This means that the small, seemingly unlikely, mid-16th century BC range for the Akrotiri dates cannot in fact be excluded on this basis. The ambiguity must remain until resolved by other means.
2. Both the Simon Fraser and Oxford laboratories found evidence for an older portion, or 'contaminant' in some seeds from Akrotiri (in particular, sample 2055). The significance was disputed, but its existence remains and, until explained, shown actually to be a contaminant (including via 'control' seeds from elsewhere), and quantified, this problem leaves a large question mark over all the radiocarbon evidence. The radiocarbon dates on 'contaminant'-free 'residues' are younger (although the Oxford dates are corrected from the average ¹³C value of -23‰ calculated by Simon Fraser, whereas the average δ¹³C value calculated from seeds by Copenhagen and Zürich is c. -20.6‰, which implies that the Oxford dates may be a little young), and when calibrated would favour a 16th century BC date range. For a graphic representation, see Fig. 1.
3. Τhe new series of dates, assumed to relate to a short period of time, display, like their predecessors from Pennsylvania, a worrying spread, often more than the 0.5 - 1.0 % error one might expect (cf. Burleigh et al. 1986; Hedges et al. 1989, 107 - 108, Fig. 1, 113). Among even the consistent set of Oxford volcanic destruction level (Phase 2/3, N = 8) and Copenhagen (N = 3) dates on seeds there is a slight bimodal tendency (3400s and 3300-3200s BP: cf. Fig. 1.A). This may reflect sample or laboratory problems, but may also be a result of the relatively large scatter (around the long-term trend) associated with dates on single-year (or short-lived) material (see Stuiver 1982, Fig. 6 and 7). However, until the problem of possible contamination is resolved, this issue cannot be tackled.

Weninger raised anticipation of a probabilistic, 'Archaeological Wiggle Matched', Aegean radiocarbon chronology, which will overcome all the ambiguities arising from the calibration of any single radiocarbon age range (and the reverse, a method to test the likelihood of a proposed archaeological chronological sequence). However, the above problems must be resolved first. Moreover, new series of high quality dates from sealed stratigraphic contexts for all the Aegean periods are required. The current corpus consists of dates from very different technical processes, and dates usually lacking carbon-13 normalization or alkali pre-treatment! This is unacceptable (Olsson 1987).

The pressing need is therefore for Aegean radiocarbon dates with the contextual and measurement quality to match the precision of the current radiocarbon calibration curves.

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

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