Stratified Pumice from Bronze Age Knossos

ARCHEOLOGICAL SITUATION

Whatever the size or impact of the 'Minoan' eruption of Thera, the object of historical enquiry must be to relate this event as closely as possible in time to the course of the Minoan civilization centred on Crete.

Accurate placing of this event in time is highly significant for the understanding of Minoan and thus Aegean history. In terms of the Minoan ceramic sequence (Popham 1967; Betancourt 1985) we know that Akrotiri suffered its last destruction as a complete town in Late Minoan IA (Niemeier 1980), and that the eruption products overlie the ruins. We also know that those ruins had undergone some clearance and repairs - their extent is still an open question - before the eruption (Doumas 1983, 134-135). The length of time between the destruction of the town and the eruption (and thus the question whether the latter could have caused the former) is also open to discussion, though in the excavator's opinion the time was short, not above a few months to two years (Doumas, ibid.). In Minoan terms, therefore, did the eruption take place in LM IA, or later? If later, were its effects responsible for the catastrophic destruction of the Minoan civilization throughout Crete in LM IB (Yes: Page 1970. No: Pichler and Schiering 1980)?

One approach to the problem is the analysis of pumice from stratified contexts in Crete itself, after it is first demonstrated that any such pumice is from Thera and no other Aegean source. Pumice has been found at several Minoan sites (Pichler and Schiering 1980, 26-30, for some of them). None of this material, however, satisfies both of the necessary criteria for resolution of the problem: specified stratigraphical and dated contexts (with published pottery) and published analyses.

Five of the discoveries are, however, of potential importance, but as yet require full documentation against the above criteria.

1. Nirou Khani. Conservation work by N. Platon after World War II brought to light a large deposit of conical cups, most of which contained pieces of pumice, under the threshold of a door leading into the room with the sheet-bronze double axes in the LM I villa (fullest discussion, with references, in Boulotis 1982, 158, 160). There is good prima facie evidence here for a pre-LM IB context, the villa having been destroyed in LM IB, but the finds have not yet been published, nor the pumice analysed.
2. Kato Zakros Palace. Particles of Theran tephra from the Minoan eruption were found and reported from levels in the palace stated to be of LM IA date (Vitaliano and Vitaliano 1974). Here the identification, by refractive index, appears clear. But the material is tephra particles, not pieces of pumice, and no details of or dating material for the context have yet been given. On the other hand pieces of pumice were found with conical cups and olives in the well of room XLI of the palace at the time of the LM IB destruction (Platon 1971, 196-197).
3. Pseira. From the excavations currently being conducted by P. Betancourt and C. Davaras waterwom pebbles of pumice are reported from the floor packing of Room I in Building AC (the Shrine), from the stratum beneath the level of the stone floor slabs (Betancourt and Davaras 1988, 218), these slabs having been used finally at the time of the LM IB destruction of the site. The preliminary report announces identification of the pumice by C. Vitaliano as from the Minoan eruption of Thera, on the basis of refractive index and mineralogical composition. 'The conclusion that the Theran eruption predated the Pseiran destruction in LM IB is inescapable.' (Betancourt and Davaras, ibid.) This again looks strong evidence for a pre-LM IB eruption. Publication of the analytical data by Vitaliano is therefore awaited with interest. Betancourt, Goldberg, Simpson and Vitaliano report more fully on this Pseiran pumice at this Congress. 
4. Khania. A remarkable stratum composed of pieces of pumice, 15-20 cm thick and extending over the whole of a room (2.3 X 1.3 m), was found in excavations in 1976 on the Mathioudakis plot in the city (Tzedakis 1976, 465). The excavator states that mixed in with the pumice were a few waterworn sherds from household vessels, LM IA in date. The overlying stratum was said to have LM I and LM IIIA fragments. Although no evidence is given for the LM IA dating of the sherds in the pumice, it is unlikely that the date of the pumice could be later than LM I, on the evidence of the stratigraphy here (LM IIIA post-dating the pumice level and LM II being absent) and of the LM IB date of the main destruction horizon at Khania. Thus the pumice was in situ at the latest by the time of that destruction. A sample of it has been analysed by Francaviglia (1986); values for both major oxides and trace elements show that it is Theran Bo. In other words, the evidence from Khania indicates that the Minoan eruption had occurred before the LM IB destruction there.  
5. Theran pumice with a Bo refractive index has also been found in a context reported to be LH IIA (equivalent to later LM IA and LM IB) at Nichoria in Messenia (Rapp et al. 1973).

KNOSSOS STRATIFIED PUMICE SAMPLES

To carry the subject forward stratified pumice samples were selected for analysis from the recent Knossos Stratigraphical Museum Site excavations, directed by Warren (1980-81; 1982-83). Ten pieces of pumice were chosen. Four of these come from Late Minoan IB or earlier contexts (four was the total number found from such contexts), six were from post-LM IB contexts. These six were added to ensure a representative selection from the Knossos collection and to enable comparison or contrast with LM IB or earlier pieces. Details of the samples are given in the catalogue. The hypothesis was that, if any of the four samples from LM IB or earlier contexts could be shown to emanate from the Minoan eruption of Thera, the eruption would have taken place before the date of the context, that is before the time of the LM IB destruction at Knossos. (We say before because the pieces in question could not have been airborne or washed into their find places at the time of the eruption; they could only have been carried to the site by human agency, i.e. they were in existence/the eruption had taken place before they reached their contexts.) Further, if the four samples turned out to be from Theran pumice earlier than the Late Minoan eruption (i.e. from the Middle or Lower Pumice series), then the possibility arises that the late Minoan eruption was contemporary with the LM IB destructions. This possibility would be greatly strengthened, though not proved, if the six post-LM IB samples all turned out to be from the Late Minoan eruption.

ANALYSES

Analyses of the ten Knossos samples were carried out by Puchelt at the Institut für Petrographie und Geochemie, Universität Karlsruhe, using three separate methods, XRF, INAA and ICP-MS. 12 major oxides were measured by XRF (Table 1); 2 of the major oxides (Na₂O and Fe₂O₃) and 24 trace elements by INAA (Table 2) and 44 trace elements by ICP-MS (21 of these by INAA and ICP-MS, 23 by ICP-MS only) (Table 3). The results were checked by analysing USGS reference rocks along with the samples. They were also treated as unknown. Values for rare earths from the INAA analyses are also given as chondrite-normalized values and plotted as such in diagrams (Table 4). To confirm the autonomy of the analytical part of the investigation the samples were known in the laboratory only by the letters A to J.

Before we turn to interpretation of the results Puchelt comments as follows on the degree of fit between the INAA and ICP-MS analyses (ppm) for 20 trace elements.

Cr       Values generally higher by INAA, but apart from sample A the same general trend is observable in both methods.

Co      Values the same within acceptable limits.

Ni       Only ICP-MS values are given; INAA values were subject to large standard deviations.

Zn      Values in good agreement.

Rb      Values in close agreement.

Sr      Comment as for Ni.

Cs      Values in close agreement.

La      INAA data 20-30 % higher than ICP-MS data.

Ce      Some INAA values up to 20 % higher, but most agree well.

Nd      Values mostly in close agreement.

Sm      Some INAA values up to 10 % higher, others closer to ICP-MS.

Eu       ICP-MS values are about 5-10 % higher than those of INAA. (The ICP-MS values have been corrected for interference from the barium-oxygen cluster forming from ¹³⁷Ba. Without this correction Eu values could be around 0.25 ppm too high in ICP-MS measurements.)

Gd      Fair agreement.

Tb      Some values very close between the methods, more about 10 % higher by ICP-MS.

Ho      Some values close, but most 15-20 % different.

Yb      INAA values mostly under 10 % higher, though samples B and C higher by ICP-MS.

Lu     Values in good agreement.

Hf      Most values in agreement to within less than 10 % difference.

Th      Most values in close agreement.

U      Most values in good agreement, though some much higher by INAA.

The analyses of Mn (ICP-MS) should also be of interest, since this trace element does not appear to have been analysed previously for pumice.

INTERPRETATION

• Within the sample series:

Inspection of the XRF data (Table 1) shows at once that sample B stands out clearly from all other samples with its lower values for TiO₂, Al₂O₃, MgO, Na₂O and Fe₂O₃ and its much higher value for K₂O. C stands out from all other samples with its higher values for TiO₂, Al₂O₃, MnO, P₂O₅ and Fe₂O₃ and its lower value for K₂O. B and C are thus also distinct from each other. Sample A stands out less clearly, but shows distinction from the remainder in respect of its higher values for MgO, P₂O₅ and Fe₂O₃, though its values for these oxides are still lower than those of C. The remaining samples, D-J, group well together. This is clear not only from the XRF data, but also from those of INAA (Table 2).

The distinctiveness of sample B in respect of Na₂O and Fe₂O₃ is entirely confirmed by the INAA data for these two oxides, as is that of A and C in respect of their high Fe₂O₃ values, with C much higher than A.

The distinctiveness of B from the rest of the group is also well confirmed among the INAA rare earth values for Sm, Eu, Gd, Tb, Ho, Tm, Yb and Lu. In all these the values for B are well below those of all other samples (and are also below all except C for Ce and Nd, where they are similar). Its Mn value (ICP-MS) is also well below those of all other samples. The distinctiveness of C is also confmned in its high INAA values for Sm, Eu and perhaps Tb and Lu, and its low value for Yb (though not nearly as low as B for this element). Its high value for Mn (ICP-MS) is also very noticeable. Among the INAA rare earth values sample A, apart from its differences from B and C, does not seem to be distinct from the rest of the series.

Among the ICP-MS rare earths which were also analysed by INAA sample B remains distinct in its much lower values for Sm, Eu, Gd, Tb, Ho, Tm, Yb and Lu. In sample C the ICP-MS values of rare earths also analysed by INAA offer less distinctiveness than the latter, but Eu and Tb remain comfortably higher than in all other samples.

In summary, then, we have four distinct groups, though our data might benefit from full statistical analysis.

The groups are (1) sample B, (2) sample C, (3) sample A (mainly XRF data), (4) samples D-J.

• Sample series in relation to other series:

The next step is to compare our data and groups with published analyses of Theran and other pumices.

The unavoidably late date at which this paper was written has had a major advantage. It means that in addition to older published analyses (Puchelt and Schock 1972 - Thera pumice, all series (Upper, Brown, Lower, Older) and Yali pumice; Vitaliano, Fout and Vitaliano 1978 - Thera Bo series and Akrotiri excavations samples; Schock 1977; Keller 1980 - Yali pumice; Schock and Pichler 1980 - Theran Bo series, Milos, Nisyros, Yali and Lipari pumices, and Cretan archaeological site samples; Francaviglia 1986 -Theran Bo and Bu series, Nisyros, Yali, Milos and Khania, Crete archaeological pumices) we have been able to use the abundant and very detailed analyses of (1) Francaviglia and Di Sabatino, (2) Francaviglia, (3) Vitaliano, Taylor, Norman, McCulloch and Nicholls in their papers for this Congress. Hereafter we abbreviate to FS, F and VTNMN.

(1.)     Sample B. All analyses show that this sample is not Theran. Its values fall off the cluster diagrams of FS and VTNMN. The individual values for TiO₂, Al₂O₃,and Na₂O fall outside and below known limits of Theran pumices (all series; the Al₂O₃ value is just within the lower limit of Bu) and above the known limits of Theran K₂O. The Fe₂O₃ value is well below the lower limit of the 131 Bo analyses of FS (and below the rest of their Theran analyses), but is within the range of the samples of Puchelt and Schock and those of Vitaliano, Fout and Vitaliano. Among the trace elements FS have Sr, Rb, Co, Cr and Ba in common with our analyses. B's Rb value is above and its CR value far above their upper limits.

Several of B's values in fact correspond well to published data for Yali pumice, namely TiO₂, Al₂O₃, MgO, K₂O and Fe₂O₃ among the major oxides (cf. both Puchelt and Schock 1972 and Keller 1980, Table 2, especially Yali 3). Among trace elements we note for example that the Eu and Lu values for sample B fit quite well with the Yali values of Puchelt and Schock, but do not fit at all well with their Theran values for these elements. However, Francaviglia's data allow us to compare B with both Yali and Milos. B's values in fact fit better with his values for Milos than with Yali on TiO₂ and MgO and closely with Milos and not at all with Yali on Mn, Nb and Ba. On the other hand B's K₂O value goes much better with Yali in both Keller's and Francaviglia's data than with Milos. On this evidence we suggest that sample B is pumice from Milos rather than Yali.

(2)     Sample C. When C's values are plotted on the cluster diagrams of FS and compared to the average and specific values in Tables 1, 8, 9 and 13 of VTNMN the result is unequivocal. Sample C is Theran pumice from an eruption older than the 'Minoan' Bo one. On the diagrams of FS its values go with either Bm II or Bu (Fe₂O₃/SiO₂, TiO₂/SiO₂, MgO/SiO₂, Al₂O₃)or just with Bm II (K₂O/SiO₂, Rb/SiO₂, Zr/SiO₂). In relation to the data of VTNMN C goes uniformly with Bu, but they were not analysing Bm II. Among the trace elements C's Ba values (INAA and ICP-MS) fit only with Bu values of FS; against Tables 6-7 of VTNMN the INAA Ba value lies better with the older pumices, though it is also comparable with some Bo samples; the ICP-MS value fits only with the older pumices. C's Mn value goes closely with Bu (ICP-MS, Mn being about 20 % less than MgO). The Eu values (INAA and ICP-MS) go very well with the older pumice values of Puchelt and Schock and VTNMN (Tables 8, 9, 13).

(3)     Sample A. The interpretation of sample A is problematical. When its values are plotted on the cluster diagrams of FS, which combine SiO₂ with another oxide or element, we have the following: TiO₂/SiO₂, probably Bu; Al₂O₃/SiO₂, probably Bu; MgO/SiO₂, not clear; CaO/SiO₂, not clear; K₂O/SiO₂ , Bu; Fe₂O₃/SiO₂, probably Bu; Rb/SiO₂, Bu; Zr/SiO₂, not clear. This evidence points clearly towards A as Bu pumice. However, when we compare A with the average values of VTNMN Table 1 we have SiO₂, Bu (but others of our samples would also be Bu on their SiO₂ values) or Bo on the values of Vitaliano, Fout and Vitaliano (1978, 209 and Fig. 3); TiO₂, probably Bo; MnO, Bo; Fe₂O₃, Bo; Eu (VTNMN Tables 8, 9, 13), Bo on our INAA and ICP-MS values. Among the cluster diagrams of VTNMN Nb/K₂O, Ba/SiO₂ and V/SiO₂ offer clear Bo/Bu distinctions. On the former A goes better with Bo, on Ba/SiO₂ clearly with Bo; we do not have values for V. Among the trace elements in Schock's data (1977, Table 1) Sm and Tb show clear Bo/Bu differences, in addition to Eu. A's values for all three elements against these data are with Bo (as are those of D-J, see below). A's value for Mn (ICP-MS) is Bo; it also lies well below the inferred average of Bu Mn (1100 ppm MnO less 20 %).

If, despite the combined Bo/Bu evidence in favour of Theran pumice, we consider a non-Theran source and set A's trace element values alongside those for Rb, Sr, Y, Zr, Nb and Ba in F's data, we obtain no illumination, the comparisons being either nil, many, or cancelling each other out for Milos and Nisyros.

Finally, therefore, we return to Thera. The Mn and Eu distinctions between Bo and Bu seem clear; on this evidence, supported by the other Bo links, we suggest A is Bo pumice; but others may feel the Bu links are sufficient to leave interpretation in doubt. The refractive index of A, however, is 1.511, placing it just on the upper limit of Bo and just below the lower limit of Bu (Bo 1.507-1.511, Bu 1.512-1.516, CT 2 1.511-1.515, Vitaliano et al., volume two of this Congress).

(4)     Samples D-J. Samples D-J cohere closely in the XRF data (Table 1) and in those of INAA (Table 2). If we relate J, which is archaeologically significant, to the FS cluster diagrams we find that its values accord with Bo on TiO₂/SiO₂, Al₂O₃/SiO₂, MgO/SiO₂, Fe₂O₃/SiO₂, Rb/SiO₂, and probably Bo on Zr/SiO₂. When we relate J to Table 1 (averages) in VTNMN we see that on TiO₂ and MnO it is Bo and nearest to but well below Bo on Fe₂O₃, and in no case is it Bu. When we relate D-I to VTNMN Table 1 we find again (since they cohere with J) that they are Bo on TiO₂, MnO and also Bo on the Ba/SiO₂ cluster diagram. These samples do, however, align with Bu on Na₂O (VTNMN Table 1 averages) and are nearest to Bu on the cluster diagram MgO/SiO₂. In Schock's data for Sm and Tb, D-J, as noted above, go with Bo. The Eu value (Puchelt and Schock 1972, Table 1; Schock 1977, Table 1; VTNMN Tables 8, 9, 13) is uniformly Bo on INAA for D-J and Bo for D-I on ICP-MS. J's ICP-MS Eu value, 1.08, is at the upper limit for Bo (cf. Puchelt and Schock 1972, sample 397) but well below the Bu values for all samples of Puchelt and Schock and all but one (1.07) of VTNMN. J's Mn value, like those of D-I, is clearly Bo. The evidence therefore suggests very strongly that D-J are Bo pumice. We must add, however, that the refractive index of J is 1.519 (analysed against several r.i. solutions). This is well above the r.i. of any Theran pumice. The only comparable value is 1.521 for Ischian pumice (Ninkovich and Heezen 1965, 418-9), the lower tephra (25,000 years old) of Mediterranean cores but also found on land in the Lipari Islands (Keller 1971, Fig. 6) and so theoretically available for sea transport. However, our element analyses show J to be Bo pumice, and an Ischian original can be ruled out on SiO₂ alone (Ischia about 60 wt%, Keller 1971, Table 1). We conclude that there is an unexplained anomaly in the 1.519 value for J, which should not displace the conclusion from the large number of oxide and trace elements analyses of the sample.

CONCLUSIONS

It will be seen from the Catalogue that four samples are archaeologically critical, since they come from LM IB or earlier contexts. They are therefore directly relevant to the relationship of the 'Minoan eruption' of Thera and its possible effect on Minoan civilization. These are samples A, B, C and J.

Sample C is from a Middle Minoan II context. It is therefore very satisfactory to find that it is from an older pumice series of Thera than that of the LM eruption.

Sample B dates to MM IIIB or the MM IIIB - LM IA transition. It is interesting to see that it is not Theran, but quite probably Milian.

Sample A is likely to date to the MM IIIB - LM IA transition, though the associated pottery does not rule out an LM IB date. It is therefore most interesting, not to say exasperating, that we cannot pin it down with certainty to the 'Minoan eruption' or a 'pre-Minoan' eruption pumice formation on Thera. The strong indication that it is Bo is compatible with an LM IB archaeological date (and so with a pre-LM IB eruption).

Sample J, like the post-LM IB samples D-I, is interpreted as Bo pumice. J's context is almost certainly LM IB and it cannot be later than LM IB.

With A and J as Bo pumice the 'Minoan eruption' took place before the LM IB destruction of Knossos and Crete. Precisely how long before that LM IB destruction (which at present we take as contemporaneous throughout the island) and how long after the LM IA destruction of and subsequent clearance and repairs in the town of Akrotiri the eruption occurred remains a subject for further discussion. To illuminate the question of the relationship, contemporary or sequential, of the eruption to the Cretan LM IB destruction horizon was the major object of our investigation.

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

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