Archaeomagnetic Results from Late Minoan Destruction Levels on Crete and the 'Minoan' Tephra on Thera

Samples taken from structures in different orientations (excluding kilns that may not have been heated during this specific event) have directions of magnetic remanence acquired as they cooled after this event which form two distinct groups. All have northerly declinations, but samples from central Crete (Gournia, Malia, Phaistos, Ayia Triada, Sklavokambos) have inclinations of 60.1^o while those from eastern Crete (Palaikastro, Kato Zakros, and possibly Makrygialos) have shallower values, 55.5°. The angular difference between the mean directions, 4.6^o, is statistically significant and is not less than 2.6° at a 95% probability level. On Thera, samples from a single hearth at Akrotiri, attributed to late LM IA, have a mean direction of remanence that is 8^o from the Cretan site directions, with 4^o between their respective circles of 95% confidence, of which only 1-2^o could be attributable to spatial changes of the geomagnetic field. The ancient geomagnetic field intensities for the same sites are not well defined but are similarly consistent with three different field strengths at the times of cooling.

These differences are not due to magnetic instability, nor the presence of other magnetic components. It is difficult or impossible to see how it could occur as a result of differential tilting or compaction. Systematic distortion by sample fabric anisotropy, magnetic refraction within the structures, or local geomagnetic field distortions are similarly highly improbable explanations. It is concluded that the most likely explanation for the directional differences is that the geomagnetic field had time to change between the destructions in Akrotiri, central and eastern Crete.

Similar studies of volcanic tephra from Thera, particularly those using alternating field demagnetization, can be interpreted as indicating a similar time interval between the eruptive sequences, with the destruction in central Crete corresponding with the eruption of the basal plinian ash, and the destructions in eastern Crete being synchronous with the eruption of the later base surge, chaotic and ignimbritic tephra. However, the lower degree of heating of these later tephra and their unconsolidated nature result in the directions of remanence in the base surge and overlying tephra being more scattered and complex than for the archaeological materials or underlying plinian tephra. Different forms of analysis can thus indicate different conclusions, some of which are consistent with a continuous evolution from plinian to a paroxysmal eruption with no distinct time-gap. Such an event would still be later than the last use of the Akrotiri hearth, and would be contemporaneous with the fire destructions in central Crete and preceding those in eastern Crete.

The time represented by the directional differences in Crete can be estimated from geomagnetic field behaviour during the last 2,000 years in western Europe. This suggests that the absolute limits (95% probability) for this time-gap could be more than 6 and less than 450 years. Comparison with directions of magnetization in deep-sea cores from the eastern Mediterranean suggest a time interval of approximately 50 years. More precise definition of the time-gap requires the establishment of an eastern Mediterranean geomagnetic secular variation curve for this period, such as now being attempted using the Egyptian materials. Establishing archaeomagnetic values for other Minoan destruction levels, and Cycladic, Mycenaean, Hittite and Levantine sites, could link these chronologies and also determine a more precise sequence of events in this region.

INTRODUCTION

The principles of archaeomagnetic directional dating are now well established (Thellier 1939; Aitken 1974; Tarling 1975, 1983) and will only be briefly summarized. They are dependent on only two basic observations. (i) The Earth's magnetic field gradually changes in direction, and (ii) magnetic particles within heated archaeological or geological objects can retain a record of this direction from the time that they cooled. This means that when a stable magnetic direction in samples from fired archaeological or geological structures has been isolated, then the original direction of the geomagnetic field at the time that they cooled can be retrieved. As the geomagnetic field direction is uniform over several 100 km, the magnetization of individual oriented samples can be used for reconstruction purposes (Burnham and Tarling 1975) or the average direction of several samples can be used for dating the structure either absolutely, by reference to other dated field directions, or relatively, i.e. establishing whether two or more structures have different geomagnetic field directions in which case they must have acquired their original cooling magnetizations at different times. However, while the basic principles are simple, there are, as with all dating techniques, some problems and these will be discussed more fully following a description of the observations on Theran and Cretan materials.

The archaeomagnetic data from Thera and Crete (Downey 1983) have already been summarized (Downey and Tarling 1984a, 1984b), but were not published in detail. The individual archaeological site results will therefore be described first, followed by a description of the results from studies of the tephra on Thera. Both sets of observations will then be discussed to evaluate the reality or not of the proposed differences in direction and palaeointensity of the geomagnetic field at the time of the LM I destructions and the 'Minoan' eruption of Thera.

ARCHAEOMAGNETIC OBSERVATIONS

• (a) Archaeological sites on Crete and Thera (Fig. 1):

(i) Thera.

Only one site was samplable at Akrotiri (Thera). This comprised a small hearth in Area 1 (Arvaniti 1, α1) - one of the first areas excavated. The archaeological age of the hearth is not clear although pithoi in the room are thought to be of LM IA design. It seems reasonable to assume that this hearth was in use until the destruction of the city prior to the volcanic eruption. As rubble below this level includes LM IA pottery, this is generally thought to be due to earthquake activity in mid to late LM IA times, particularly as there is evidence for clearance and possibly rebuilding prior to the first ash fall (Doumas 1974, 1978).

Orientation was difficult because the roof prevented the use of a sun compass and this may account for some of the scatter in directions, α₉₅=3.0^o to 3.6^o (although this magnitude of scatter is typical for such structures). Alternating field (AF) demagnetization (Fig. 2a) showed virtually no change in the direction of remanence (17 samples; mean directions - initial 9.7°, 58.1 °, α₉₅= 3.0°; after partial AF 10.6°, 57.3°, α₉₅= 3.6^o). The past field strength, using 6 specimens, was estimated to be 51.0 ± 4.9 μT.

(ii). Crete.

On Crete, a total of 10 separate localities (18 sites/structures) were studied, comprising isolated villas, town sites and palaces. A range of materials was sampled, although the vast majority were of clays (mostly mud brick) that had been very strongly heated (> 600^o C) during the Late Minoan I fire destruction of these localities. These fired clays were found to be ideal materials for archaeomagnetic analysis as they contained virtually only one magnetic component - even viscous components were so weak that there was virtually no difference between the direction of remanence before and after partial demagnetization (Fig. 2 b, c). Thus thermal demagnetization up to c. 580° C and alternating magnetic field demagnetization to c. 100 μT revealed very high stabilities of remanence (stability indices mostly > 5 - Tarling and Symons 1967) and clear linearities (diagonal angles generally ≤ 2° - Kirschvink 1980) even if the initial direction was included in the analysis. However, other materials, such as charcoal ash and heated sandstone horizons, showed complete consistency with the directions isolated in fired mud bricks from the same locality.

The following summary is given in alphabetical site order:

     Ayia Triada.      Samples were taken from the east wall, floor and two other walls in the Palace and also from a kiln. The kiln had statistically identical directions both before and after partial AF demagnetization (N=5; initial 3.5°, 60.8°, α₉₅=4.7°; after partial AF 4.5°, 61.3°, α₉₅=4.3°) and the estimated past field strength was 65.8 ± 4.1 μT, based on 6 samples. The Palace samples also showed high consistency before and after demagnetization (initial mean direction 354.4°, 62.6°, α₉₅= 5.9°, N=20; after partial AF 353.8°, 60.1°, α₉₅=2.3°, N=17) and a field intensity of 68.1 ± 2.0 μT, N =6).

The last firing of the kiln and its relationship to the destruction of the Palace are not known but Levi has suggested the destruction was in LM IA times (Warren 1980). The archaeomagnetic data suggest a similar age for the last firing of the kiln and of the Palace destruction, but differences of a decade or so cannot be excluded (see later discussion).

     Gournia.      The age of the fire destruction of this small industrial town is LM I, but the precise age (A or B) has not been established. Samples were taken from mud-brick walls in House A c Room 18. The initial mean direction (349.2°, 63.7°, α₉₅= 11.2°, N=30) was not well defined and, after partial AF demagnetization (Fig. 2b) and the removal of 6 deviant samples, became more northerly and shallower in direction but was much better defined (356.1°, 59.5°, α₉₅= 1.0°, N=24). The past geomagnetic field strength, based on 6 samples, was estimated to be 63.4 ± 0.8 μT.

     Kato Zakros.      Burnt mud bricks from the Palace walls and floor were sampled extensively and showed high stability to AF demagnetization, with little change in direction, although the elimination of 6 somewhat aberrant directions considerably reduced the scatter (initial mean direction 356.1^o, 56.9^o, α₉₅ = 7.3^o, N=55; after partial AF 355.8^o, 54.8^o, α₉₅= 1.9^o, N=49). The past geomagnetic field strength was estimated to be 60.7 ± 2.4 μT based on 6 samples. The archaeological date for the Palace destruction may have been some 50 years later than that of Ayia Triada (Warren 1980) but pumice from Minoan Thera has been found in reputedly sealed sites prior to LM IB times as both marine and floral styles of pottery have been found; however, there is some dispute as to whether the sites were truly sealed.

     Knossos.      It was not practicable to collect from most of this location other than at kiln 2, recently excavated as a part of the Stratigraphic Museum extension (Warren 1981), which clearly pre-dates LM IB pottery found in some of the channels, but by an unknown time. These results are described more fully elsewhere (Tarling and Downey, in press). The initial mean direction of remanence (354.7°, 59.3°, α₉₅= 3.1°, N=26) was identical to that after partial AF treatment (355.2°, 60.8°, α₉₅= 1.6°, N=25) and the ancient field intensity was determined as 65.1 ± 6.7 μT.

     Makrygialos.      Burnt mud bricks from the walls and floor of this villa, together with charcoal ashes, showed slightly scattered directions before partial AF demagnetization (356.0°, 54.8°, α₉₅= 8.3°, N=35) but very consistent directions afterwards (354.7°, 57.5°, α₉₅= 1.4°, N=32) and a past geomagnetic field intensity of 61.9 ± 5.7 μT. The precise age of destruction is not clear, although of LM I age.

     Malia.      This extensive site was sampled at four different locations comprising:

(1).     Area 13 of the Palace from which 20 samples of burnt, vertical mud brick walls were taken. These provided statistically identical directions before (355.0°, 59.5°, α₉₅= 1.4°, N=20) and after (356.1°, 59.6°, α₉₅= 1.4°, N=20) partial AF demagnetization and an estimated past geomagnetic field strength of 61.7 ± 2.6 μT. Pelon (1970) considered that the destruction of this palace occurred in LM IA times, but may well have overlapped into LM IB times.

(2).     Burnt mud brick walls were also sampled at Maison α and yield an initial mean direction of 350.5°, 64.2°, α₉₅= 13.5° (N=24) which, after partial AF demagnetization, changed slightly but became much better defined (356.6°, 59.9°, α₉₅= 3.2, N=22). The destruction is assumed to be the same as for the rest of Malia, i.e. most pottery is of LM IA style, and the general opinion is that destruction was in early LM IB times.

(3 and 4)     A burnt horizon containing shards, ash etc., was sampled in Quartier E room 3 coupe 2, but it was not clear whether this horizon was completely in situ since the original destruction. The initial directions were very scattered (1.5°, 44.9°, α₉₅= 24.4°, N=34) and while these were mostly very stable, many specimen directions were clearly deviant even after partial AF demagnetization and more than half of the specimens had to be eliminated (6.4°, 59.6°, α₉₅= 7.4°, N=15). Clearly, any interpretation based on this result alone must be treated circumspectly. A hearth was also sampled in Quartier E, room 3, although, as with the burnt horizon, the foundations were not fixed and different parts of the structure could have moved since they were last fired. The initial scatter was high (286.7°, 59.3°, α₉₅= 17.5°, N=19) and, although stable, remained high after partial AF demagnetization. The removal of two thirds of the most deviant samples provided a mean direction of 327.8^o, 62.8^o, α₉₅= 12.2^o (N= 8) but this result must be treated even more cautiously than for the burnt horizon. The destruction in this location is assumed to have occurred at the same time as the rest of Malia.

     Palaikastro.      This locality, which originally comprised a large town, is known to have been occupied until at least LM IIIA times, although it suffered major destruction in LM IB times. Samples of burnt mud bricks were taken from floors and vertical walls in Block N, and the floor samples included black charcoal ashes. The initial directions were all similar (353.3°, 55.4°, α₉₅= 4.1°, N=59) and virtually identical after partial AF demagnetization (355.1°, 55.1°, α₉₅= 1.8, N=59 - Fig. 2c). However, the directions isolated in the floor samples were slightly shallower and more westerly (357.1^o, 53.7^o, α₉₅=2.2^o, N=38) than those from the walls (353.6^o, 58.2^o, α₉₅=1.6^o, N=20). The geomagnetic field strength, based on 6 specimens, was estimated to be 53.5 ± 2.9 μT. An empty pottery kiln of uncertain age was also sampled at Palaikastro. This had stable remanences (initial mean direction 359.7°, 48.4°, α₉₅= 3.0°, N=18) essentially identical after partial AF demagnetization (0.6°, 47.1°, α₉₅=3.2°, N=18). The direction of remanence is so different to that for other LM I sites that it is suspected that this kiln belongs to LM II or LM III times, although such interpretation awaits the definition of the geomagnetic field directions for other periods.

     Phaistos.      Burnt mud bricks were sampled at three localities: (1) Area 101 Archive (site of the discovery of the Phaistos disk), (2) North of the Central Court wall, and (3) the north wall of room 70, north-west of the Central Court. The directions of remanence were essentially identical before (356.6°, 63.1°, α₉₅= 2.7°, N=22) and after partial AF demagnetization (356.6°, 61.1°, α₉₅= 1.7°, N=17) and the ancient geomagnetic field strength, based on 3 samples, was estimated to be 67.3 ± 6.6 μT.

The walls of a kiln which was part of the Old Palace - considered to have been last used at the end of the First Palace Period, MM IIB (or possible LM I) - was also sampled but had very different directions when compared with other LM I sites both before (7.7°, 40.4°, α₉₅= 7.4°, N=6) and after partial AF demagnetization (6.0°, 41.5°, α₉₅=3.5°, N=5). On this basis, combined with the limited archaeological control, it is considered that this kiln is not of LM I date, but more precise age assessment required data from both younger and older Minoan sites.

     Sklavokambos.      Burnt mud bricks were sampled at this villa and yielded an initial mean direction of 355.8°, 54.4°, α₉₅=15.3° (N=23) which improved precision after partial AF demagnetization and removal of 7 outlying specimen directions (354.3°, 59.1°, α₉₅= 3.9°, N=16). The ancient field strength was determined to be 69.2 ± 2.7 μT.

     Tylissos.      Only one un oriented burnt mud brick was obtainable from House A and this provided an estimate of the ancient field strength of 81.1 ± 3.6 μT (N =5).

• (b) Magnetic studies of the volcanic rocks on Thera:

Previous studies (Tarling 1978; Wright 1978) had shown that the black and red andesitic/basaltic volcanic 'bombs' each had magnetically stable directions of remanence, but the directions of different blocks were randomly oriented, i.e each bomb had acquired its remanence prior to emplacement and had not subsequently been reheated (except possibly in their outer 1-2 mm but attempts to measure the magnetization of this thin skin were unsuccessful at that time). Only two tephra blocks had been sampled by Tarling (1978) as the study was then restricted to large volumes of rock that could be drilled in the field. Both of these blocks had mutually consistent directions of remanence, but it was not felt that any reliance could be placed on only two samples as geomagnetic field indicators. However, it was on this basis that later sampling by Downey, in 1980 and 1981, was concentrated on the light (colour and weight) volcanic tephra and a sampling procedure originally developed for archaeomagnetic work was then adopted whereby small (< 1-2 cm), irregularly shaped pumice samples were oriented and collected using the disk method (Tarling 1983). The remanence of these samples could then be dissected by alternating magnetic fields, without further cutting, and some samples could be reoriented within pyrex cylinders for thermal analysis. The results of the tephra studies are described below in chronological order, commencing with the oldest deposits, following comments on some of their common characteristics.

The results of studies of further collections of the lithic xenoliths are described in more detail elsewhere (Tarling and Downey, in press) and are not repeated here.

The magnetic stability and linearity characteristics of the remanence of the tephra samples (Fig. 3) were less than for the archaeological materials, but this appears to be mainly due to two factors.

(1)     The narrower, lower temperature range over which the remanence direction had been acquired, as many of the tephra clasts were emplaced at temperatures of only some 300-350° C and viscous components could be dominant at temperatures below 150° C. Both alternating magnetic fields and thermal methods of demagnetization were used to isolate the primary component. The initial summary (Downey and Tarling 1984a, 1984b) was based entirely on the results of alternating field demagnetization, which represented the vast bulk of the study, but greater attention is paid here to thermally demagnetized results.

(2)     The unconsolidated nature of the tephra also meant that it was difficult to orient the individual tephra without physical disturbance, and occasional samples had directions which were so deviant from the others that they resembled the random patterns observed in the lava blocks, indicating that some of the tephra pellets were similarly emplaced after they had already acquired their remanence (Tarling 1978; Tarling and Downey, in press). It was thus necessary to exclude all sample directions that were more than 20° from the mean direction, irrespective of their stability or linearity of remanence.

The magnetic carrier of the remanence in all tephra deposits is somewhat impure titanomagnetite. No haematite could be identified on Curie point, isothermal remanence, microprobe or X-ray analyses, although traces of maghaemite were occasionally present. Optical examination shows large globular titanomagnetite grains, 40 μm to 3 mm in diameter, in intimate contact with pyroxene and feldspar phenocrysts and with a glassy matrix, some of which invades the larger titanomagnetite grains and contains unidentifiable fine specks of metallic reflectors. Microprobe analyses of the larger titanomagnetite grains suggests a uniform composition of Fe_3.6 Ti_0.4O₄ although the oxidation state was variable, but with no consistent trend throughout the tephra sequence. This composition is consistent with the dominant Curie point observed, c. 425° C, while a higher Curie point, mostly around 520^o C but up to 550^o C, is probably related to the fine grains within the glassy matrix which would therefore have a composition approaching that of pure magnetite. It is this higher Curie temperature mineral that provides the maximum temperature for estimations of emplacement temperatures of the different tephra deposits.

     (i)      The basal (plinian) ash. The lower (plinian) ash was sampled at Fira Quarry (39 samples), Akrotiri (16), Oia quarry (7), Kamari Gorge (close to Akrotiri - 5), and 4 from roadside exposures between Thera and Oia. The vast majority of these samples were white pumice clasts, some 3-4 cm in volume, and larger samples (6-8 cm) were halved to provide specimens for both thermal and alternating field demagnetization. Twenty-seven samples were thermally demagnetized (11 from Fira quarry; 11 from Oia quarry) in order to determine their thermal stability and probable emplacement temperatures. The samples from Fira quarry showed high to very high stability from above 50-100° C to temperatures between 450 and 520-550° C, with the majority showing random directions developing shortly above 500° C (Fig. 3a shows an example of stability and linearity to c. 350° C, which then becomes magnetically unstable at higher temperatures.) As the magnetic remanence is carried by titanomagnetite with a maximum Curie temperature of 520-550° C, the onset of instability just over 500° C is interpreted as indicating a clast emplacement temperature of this order. The samples from Akrotiri and nearby Kamari gorge were of apparently identical composition, form, etc., to those at Fira quarry, but became randomly magnetized at temperatures above 250-350° C, suggesting emplacement temperatures some 250^o C lower than at Fira quarry which is some 7-8 km nearer to the vent. In view of the short range over which the remanence was acquired, particularly near Akrotiri, and the physical instability of the sections, many samples had to be rejected on the basis of inadequate definition of the magnetic vector between 100-150° and 350° C or clear deviations from the grouped directions (usually > 40° from the mean direction) which are thought to represent epiclastic re-working or post-emplacement movement (Tarling and Downey, in press). Only 37 samples were considered to have satisfied these criteria during alternating field demagnetization, and only 14 'remaining' vectors and 18 'removed' vectors were acceptably isolated during thermal demagnetization. These were all combined to give a mean direction 359.8°, 59.3^o (N=69, k=39, α₉₅ = 2.8°) involving samples from all localities except the Thera-Oia roadside exposure for which no consistent samples could be found.

     (ii)      The base surge deposits. The base surge was sampled at Fira quarry (44 samples), Oia quarry (11), Akrotiri (5) and Kamari Gorge (5). Most samples were white tephra clasts, 6-8 cm in volume which were split for both thermal and AF demagnetization. Thermal demagnetization (27 Fira quarry, 4 Oia quarry, 4 Kamari Gorge) showed variable stability (Fig. 3b). The loss of consistent directions in samples from Fira quarry was usually between 300° and 450° C in the higher parts of the base surge and 450-520° C in the lower parts. The thermal behaviour of samples from Oia suggested clast emplacement temperatures around 450-500° C, and Kamari Gorge samples indicated 200-250° C emplacement temperatures. In view of the generally low range over which the vector was definable, many samples showed apparently poor stability characteristics. The mean direction is based primarily on samples from Fira quarry and included both alternating field and components removed thermally to provide as mean direction 359.1°, 55.7° (N=70, k=28, α₉₅=3.2°).

     (iii)      The chaotic ashes. These were sampled at Fira (21 samples) and Oia quarries (16), as well as at Kamari Gorge, although the latter did not provide any usable directional data. Clast emplacement temperatures were somewhat low, being mostly between 250° and 300° C at Kamari Gorge, and between 300° and 350° C at both quarries, but higher emplacement temperatures were also recorded suggesting variability in the temperatures of individual clasts as they were deposited. Thus the minimum clast emplacement temperatures probably represent the maximum tephra environmental temperature at the time of emplacement. However, the low range of blocking temperatures available and the physical instability of individual clasts meant that the final mean direction (358.5°, 57.4°) was not well defined (N=34, k=7, α₉₅ = 10.0°) despite being based on alternating field and thermal components, both removed and remaining.

     (iv)      The un welded ignimbrites. These were sampled at two localities just north of the desalination plant at Monolithos (28 samples) and between Akrotiri and Exomiti (40). All the tests for clast emplacement temperatures indicated values ≥ 400-450° C, which is consistent with their unwelded nature. However, the directions obtained were commonly poorly defined (Fig. 3c) and scattered. The most consistent data appeared to come from thermal demagnetization for which analysis of components removed and remaining gave a mean direction of 1.4°, 59.7° (N=31, k=10, α₉₅ =8.8°).

DISCUSSION

• (a) Comparison of site vectors:

     (i) Archaeological sites.

It is inappropriate to include remanence data from the kilns in any appraisal of the relationships between remanence vectors of the LM I destruction levels as these could have been last fired either before or after the destruction of the settlements. However, they do provide further evidence of the reality of the observed differences and suggest the probable trend in the change of direction of the geomagnetic field over these periods. The directional data from all other archaeological sites (Table 1 and Fig. 4) are clearly distinguished from that of the present geomagnetic field, thus confirming that any viscous components acquired during the last few years have been effectively removed during the early stages of demagnetization. It is also clear that the mean direction for the Akrotiri hearth is quite distinct from that of all destruction levels on Crete. The Cretan site directions are all broadly similar, being all approximately 5° W in mean declination, but showing a range in mean inclinations between 55° and 62°. However, the mean directions for Sklavokambos, Ayia Triada, Malia Palace, Malia town, and Gournia are all statistically identical to each other and centred on 60° inclination. Phaistos has a 2° steeper inclination than these other localities, but the overlap of their circles of 95% confidence suggest that this difference is not statistically significant. These sites in central Crete have a mean direction of 355.6°, 60.1° (120 samples, k=138, α₉₅= 1.1°) - including Sklavokambos. The mean directions of the eastern Cretan sites, Kato Zakros and Palaikastro, which are statistically identical to each other and have a combined mean of 355.7°, 55.5° (106 samples, k=155, α₉₅ = 1.1^o). This is statistically distinct (> 95% confidence level) from those of the central Cretan sites (other than the poorly defined Sklavokambos direction). The remaining site, Makrygialos, regarded here as being in eastern Crete, has an intermediate mean direction that is more closely associated with Palaikastro and Kato Zakros than with the central Cretan sites but cannot be confidently assigned to either group.

The data from archaeological materials therefore indicate at least three different geomagnetic field directions. The hearth at Akrotiri, probably having been last used in mid/late LM IA times, would be expected to differ from the Cretan destruction levels which are mostly ascribed to the end of LM IA or LM IB times. However, two different field directions appear to be clearly identified in Crete, one associated with the sites in central Crete and one with sites in eastern Crete. These directional differences are also supported by the tendency for lower ancient geomagnetic field strengths in the central Cretan sites (Table 1), but it is not considered that these determinations are individually sufficiently reliable to be considered as full confirmation of such differences.

     (ii) Volcanic tephra.

The magnetic characteristics of the volcanic tephra make their archaeomagnetic record much less precise than that for the archaeological materials. This mainly reflects that the tephra have mostly cooled in situ from much lower temperatures than the archaeological materials, and the individual volcanic clasts are physically loose within the exposures, making it difficult to orient them without movement, and many also appear to have moved prior to collection due to post-depositional settling or even re-working (Tarling and Downey, in press). It is noticeable, however, that the sample directions that were rejected on the basis that they were > 20° from the mean direction were almost all strongly deviant, virtually all being more than 50° from the mean. This suggests that the error due to orientation difficulties was probably less than 2° and that it is the epiclastic movement of tephra clasts that is the main source for the observed scatter in directions (Tarling and Downey, in press).

The plinian (basal) ash remanences are certainly the most reliable of the results from the tephra and this layer has a well-defined overall mean direction, based on both thermal (components remaining and removed analyses) and alternating magnetic field demagnetization. The base surge is more difficult to analyse and its thermal components remaining are essentially undefined because of the low emplacement temperature. Combined component analyses, using the thermal components removed and the alternating field demagnetization values, yield a mean direction of 359.1°, 55.7^o (N=70, k=28, α₉₅ =3.2^o) which is not very different from the previous analysis by Downey (1983) based on alternating magnetic field observations only - 0.2°, 56.0° (N=61, k=45, α₉₅ =2.7^o).

Neither mean direction can be considered as distinctly different, at a 95% confidence level, from that of the underlying plinian ash as their circles of confidence very slightly overlap, but a difference is clearly indicated at a lower precision level. The mean direction of the chaotic upper ash is the most difficult of the volcanic tephra to assess, probably because of a large quantity of reworked material and their low clast emplacement temperatures (Downey and Tarling, in press). In the previous analysis (Downey and Tarling 1984a, 1984b), the unreliability of these observations was considered too great to merit further assessment. However, two selection procedures have been attempted since, both incorporating the results of alternating magnetic field component remaining analyses, but the first assessment including thermally removed components (N=30, 359.2°, 54.6°, k=7, α₉₅ =11.1°) and the second including both thermally removed and remaining components (N=34, 358.5°, 57.4°, k=7, α₉₅ = 10.0°). While both mean directions are statistically identical, this mainly reflects their poor definition and this also means that neither mean direction can be statistically distinguished from the mean directions for either the plinian or the base surge tephra. The unwelded ignimbrites are similarly difficult to interpret unambiguously as so few data meet the requirements of adequate definition and consistency. This is partially due to the surprisingly high content of reworked clasts, c. 20%, yet the ignimbrites would normally be expected to provide more consistent data than any of the underlying tephra as their emplacement temperatures are higher. However, as with the chaotic ashes, it is felt that the poorly defined mean direction, 1.4°, 59.7° (N=31, k=10, α₉₅ =8.8°) should be used in evaluating the overall mean direction for the entire tephra sequence.

     (iii) General comments.

Any analysis in which data have been excluded must clearly be interpreted cautiously and the grounds for such exclusion must be objective. In this study, three main rejection criteria have been applied.

(a) The individual directional measurement must have a repeatability within at least 10°.

(b) The component isolated by partial demagnetization must have a stability index (Tarling and Symons 1967) of ≥ 1.0 and linearity defined by a diagonal angle (Kirschvink 1980) ≥2° for archaeological materials and ≥ 5° for tephra specimens.

(c) The individual sample direction must lie within 10° of the mean direction for archaeological sites and 40° for tephra sequences. Such criteria are, in standard palaeomagnetic analyses, fairly minimal criteria and it is thus reasonable to consider that any statistical differences between the mean directions of any sample groupings must have some physical explanation and are not associated with measurement or collection errors. Similarly, the departure of all the mean directions from the present geomagnetic field and that of the axial geocentric dipole field suggests that any viscous remanent effects have been essentially removed from the magnetic components used. This is also indicated by the absence of any 'stringing' of the individual specimen directions at any given site.

• (b) Potential physical causes for mean vector differences:

As the difference in vector direction for the archaeological sites is statistically so clear, the possible physical causes for such differences will mainly be considered in terms of these materials.

     (i) Geomagnetic.

As the Earth's magnetic field varies laterally over the Earth's surface, differences in declination and inclination could be expected with the inclination tending to be on average some 1.1° shallower in Thera than in Crete (the present difference is 1.5°) assuming the standard axial geocentric dipole model for the average geomagnetic field. However, while this explanation may provide a reason for small differences between Crete and Thera, the individual areas involved are far too small to expect detectable differences if they had all been magnetized at the same time, i.e. in the same geomagnetic field. It is possible that the volcanic eruption was associated with local changes in the geomagnetic field, but these are usually of the order of only 0.01% of the normal geomagnetic field (less than the diurnal variation or even magnetic storms) and could not be expected to affect the direction of remanence acquired - even on Thera itself. As the civilization was a Bronze Age culture, it is improbable that iron objects were present and impossible to believe that such local anomalies would coincide over such a large area.

(ii) Subsequent differential movement of the sites.

If all the Cretan destruction levels had been magnetized at the same time (and thus initially had the same direction of remanence), then a subsequent 5° northerly tilt of central Crete could account for the present difference between the directions isolated in central and eastern Crete. As Crete is some 50 km north to south, such a tilt would give rise to a difference in elevation of well over 4,000 m between the Minoan coastlines on the two blocks. While minor differential motions have taken place, as indicated by the present-day location of Minoan and Roman harbour installations (Flemming 1978), the angular displacements, if uniform, would still correspond to only 0.01° per 10 m difference in elevation between the north and south coasts of central and eastern Crete. Differential settling could similarly account for the deviation in any one site, but an identical settling direction and amount in all central Cretan sites is so improbable as to be impossible. In all cases, the walls which were sampled appeared to be essentially vertical and a systematic deviation from the vertical would have been clearly visible

(iii) Magnetic anisotropy and 'refraction'.

The direction of remanence of an individual specimen, or specimens from the same position, can be distorted away from the ambient field at the time the magnetization was acquired by either shape or crystalline magnetic anisotropy. However, the heating and cooling of specimens of both archaeological and tephra specimens in a known field of 60 μT showed no detectable deflection, i.e. no thermal remanent anisotropy and this physical effect is therefore discounted. Magnetic 'refraction' (Harold 1960; Weaver 1961; Aitken et al. 1971; Abrahamsen 1986) is more difficult to evaluate as the physical causes are unclear (Tarling et al. 1986). However, the intensity of natural remanences was not particularly high, generally of the order of 10 mA/m for the tephra and mostly 100 to 1000 mA/m for the burnt mud bricks. The low intensity of remanence of the tephra would exclude significant 'refraction' effects but these could be present in the more strongly magnetized archaeological specimens. However, the internal agreement within each site is extremely high, yet sampling had deliberately involved floors and walls in different orientations to test for such effects, which suggests that any 'refraction' present must have been uniform over the entire sampled area, commonly several 100m² . Only in Palaikastro could any difference be detected between the directions in the walls and floors and it is probable that such differences arose from differential settlement of the floors rather than 'refraction'.

(iv) Conclusion.

The difference between the mean direction for the Akrotiri heart and the Cretan destruction levels is clearly attributable to a difference in the age of the firings and this is consistent with the archaeological evidence for a time-gap between the seismic destruction of the settlement on Thera and the contemporaneous eruption of the volcano and fire destructions in central Crete. The only factor that seems capable of accounting for the observed difference in inclination between the central and eastern Cretan destruction levels similarly seems to be an age difference during which the geomagnetic field became c. 5° shallower, with little change in declination, and possibly increased in intensity by more than 10 %.

The situation with regard to the existence, or not, of an age difference between the plinian and base surge deposits must remain unclear on currently available magnetic observations. Taken at their face value, there is some indication for a time-gap between the two events - as originally proposed (Downey and Tarling 1984) - particularly as the difference is remarkably identical to that observed between the central and eastern Cretan sites. However, these tephra directions are not statistically differentiated at a 95 % confidence level and if the poorly defined mean directions for the overlying chaotic ashes and unwelded ignimbrites are included, then the difference is even less well-established because of the decrease in the overall precision and the somewhat steeper average direction for the chaotic ashes and ignimbrites.

The rate of change of the geomagnetic field at the time of the LM I destructions in Crete is not known, so that the chronological significance of the difference of 4.6° in inclination (2.4° between the associated 95 % probability error circles) can only be assessed in terms of predicted geomagnetic field changes based on observations in Europe during the last 2,000 years (Tarling 1988, 1989). The average rate of change of the field in Britain has been 0.05/year, i.e. 4.6° would correspond to 92 years and 2.4^o corresponds to 48 years. However, the slowest rate of change would suggest 460 and 240 years and the fastest rate would correspond to 13 and 6 years respectively. As the angular differences between the Akrotiri hearth mean direction and those of Crete, or of the base surge deposits, are similar, c. 8° between the mean directions and c. 4° between the 95 % confidence levels, time intervals of similar magnitude can be postulated between the last use of the hearth, the plinian eruption and also the Cretan fired destruction levels.

CONCLUSIONS

At this stage, there appears to be distinct archaeomagnetic evidence for a difference between the ages of destruction of LM I sites in Akrotiri (Thera), central and eastern Crete. There is similarly distinct evidence for the contemporaneity between the Cretan destructions and the volcanic eruption of Thera, although the data are not sufficiently precise to exclude differences of a few years, i.e. the destructions were not necessarily directly caused by the eruption. The hearth at Akrotiri (Thera) was not heated by the tephra that fell on this site, and was clearly last used well before the ash fall - probably shortly before or even during the seismic destruction (possibly 'slow' earthquakes - bradyseisms - as recently affecting Pozzuoli near Naples) of this settlement. The identity between the mean directions of the plinian eruption and the destructions in central Crete certainly would seem to rule out any direct volcanic cause of the fired destructions on Crete as little or no ash fell on this area and any volcanic ashes would, in any case, have been cold and incapable of directly causing fires. A seismic cause for these destructions is thus strongly indicated with the implication that the intense firing is possibly associated with the lighting of previously stored olive oil by lamps or fires spread as a result of the seismic event which must have been most intense within central Crete.

The timing of the second phase of the Thera eruption must be regarded as more problematical than was initially thought. The magnetic mineralogy suggests that there was little or no change in the redox conditions in the parent magma for both plinian and later phases, but that would also constrain models which involve changes in the nature of the volcanicity due to the intrusion of significant quantities of water into the magma chamber. However, if the volcanic explosion was due to convective overturn within the magma chamber, then this could have resulted in the rapid expulsion of all previously formed tephra, thus accounting for the large quantities of reworked material in the chaotic and unwelded ignimbrites (Downey and Tarling, in press). However, such re-working may be a common feature of all such poorly consolidated pyroclastic deposits.

Absolute archaeomagnetic dating still requires a geomagnetic secular variation record to be established on already well-dated materials. Such work is in progress in Bulgaria (Kovacheva, pers. corom.) and is being planned in Egypt. Studies of the tephra in deep-sea cores from the Mediterranean (Downey 1983) confirm that the geomagnetic field was shallowing with time, but estimates of deposition rates are too imprecise for dating. It is, in any case, vital that other archaeological sites within the eastern Mediterranean, particularly with a well-established stratigraphy, are examined magnetically to determine chronological correlation between older and younger Minoan sites and their relation to Cycladic, Mycenaean, Hittite and Levantine sites. Such studies can then link these largely independent chronologies and allow a more precise sequence of events to be determined throughout this region.

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

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