Radiocarbon Dating
Carbon-14 atoms naturally combine with oxygen atoms present in the atmosphere to produce carbon dioxide. This, like carbon-12, is absorbed by plants through the process of photosynthesis and reaches other organic material through the food chain. Carbon-14 differs from the more common carbon-12 isotope in that it is radioactive, decaying at a constant, measurable rate. This causes the alteration of carbon-14 into a new element, nitrogen, and it is this rate of decay that can be measured. Whenever organic material dies it no longer absorbs atmospheric chemicals. This means the amount of carbon-14 remaining in a sample can indicate how long the period of decay has been ongoing, in other words, how much time has elapsed since the sample died.
Radiocarbon dating originally suffered from the forced assumption by Libby (1963; 1965) that the amount of radiocarbon present in living organisms has been constant throughout time. Since his time, however, it has been found that the amount of radiocarbon-14 being produced in the Earth's atmosphere has been effected by periodic fluctuations (Renfrew 1990, 77). Thus, the method of radiocarbon dating required calibration against a standard. In this case the best source of precisely dated organic samples was dendrochronologically dated wood. Suess (1965; 1967) was able to present a calibration graph based on the annual growth-rings present in Californian bristlecone pines and this has subsequently been fine-tuned by Stuiver and Becker (1993) and Stuiver et. al. (1998). Unfortunately, due to these natural fluctuations in atmospheric carbon-14, the calibration curves do not allow a precise resolution of radiocarbon determinations (Renfrew 1990; Baillie 1995; Renfrew and Bahn 1997), meaning that radiocarbon dating can only provide date ranges rather than precise historical dates
It should also be stressed that earlier published radiocarbon dates may carry with them a certain degree of inaccuracy. The processing of contaminated samples and the use of commercially influenced laboratories have both served to provide a lot of radiocarbon dates that are practically worthless and meaningless. It is most important to use laboratories that can produce high-precision dates and where extensive chemical pre-treatment of samples from known provenance is routine. Whenever samples associated with the eruption of Thera are submitted to laboratories that fulfil these criteria the results do not rule out the possibility that Thera erupted late in the seventeenth-century BC, but due to the 'flat spot' on the calibration curve at this time, a late sixteenth-century date is also possible (Baillie 1995; Housley et. al. 1999)
Looking back at the early days of carbon-14 being used as an historical dating technique, contaminated samples giving inaccurate results and the then unrealised fluctuations in carbon-14 entering the Earth's atmosphere made this technique suspect at best and totally unreliable at worst (Daniel 1959). The upshot of this is that the spread of dates produced by the radiocarbon dating technique may have led to the false association of Aegean and Egyptian archaeological contexts. This problem is now being redressed and the present situation is adequately summed up by Manning who states that "C-14 may be high-precision dated by the best laboratories to <±20 C-14 years, with the proper pre-treatment to remove contaminants, proper allowance for isotopic fractionation, appreciation of inter-laboratory offsets, and so on. The need for the acquisition of appropriate samples from high-quality and specific contexts is now appreciated by most archaeologists, and dates are usually relevant to some event or horizon" (Manning 1996, 29).
Calibration and 'Flat Spots'
The 'flat spot' which appears on the various calibration curves (allowing either a seventeenth- or early sixteenth-century beginning for the Aegean Late Bronze Age) gives rise to chronological difficulties with the LM IA ceramic phase, as well as the dating of the eruption of Thera, in the form of a calibrated date range which is sufficiently wide to be of little value at a time when particular precision is called for. An example of the wide calibrated date range may be seen with the carbon-14 determinations obtained from three short-lived samples (charred barley) found in a LM II context vessel from the destruction level of the Minoan Unexplored Mansion at Knossos (Popham 1984). (This LM II destruction level is, of course, later than the Theran eruption which occurred during the LM IA ceramic phase.) The 1993 decadal curve (Stuiver and Becker 1993) was deemed the most appropriate for the calibration of these determinations, however, the imprecise nature of calibrated radiocarbon dates was highlighted by the results which proved to be ambiguous; two fell in the mid fifteenth-century and the other in the fourteenth-century BC, while the date range at 1σ covered the period 1540-1310 cal BC (Manning and Weninger 1992, 651).
Contamination and Old CO2
Sample contamination can also give misleading results. An example of contaminated samples giving false dates was initially thought to have been observed in the analysis of pulses and grains (N = 15) originating from the LM IA (i.e. Thera) destruction level at Akrotiri and published by a team from Oxford. The uncalibrated weighted average of these determinations was 3337±16 BP (Housley et. al. 1990). Although these readings gave apparent hope to those searching for an early eruption date, it was claimed by workers from the Simon Frazer University (Nelson et. al. 1990) that an older contaminant had been present in the samples, in response to which the Oxford team re-evaluated their results to produce the Series II dates. Here, four uncontaminated samples gave the weighted average of 3299 ± 30 BP. Although this date appeared to be significantly lower than the earlier 3337 ± 16 BP, the fact that both date ranges fell on the same 'flat spot' on the calibration curve - and could both be calibrated to the seventeenth- and sixteenth-centuries cal BC - indicated that both ranges were statistically acceptable.
Nevertheless, even with these redetermined dates there are still quite a few unresolved problems regarding the application and accuracy of radiocarbon as a dating technique. As the Akrotiri samples were of grains and pulses (therefore short-lived), there is no reason to believe that higher dates were due to the 'old wood' effect as they are unlikely to have been harvested more than a year before their entry to the archaeological record. (The 'old wood' effect is caused by CO2 having been absorbed by trees at earlier stages of their lives and can be contained in organic material that has been around for a period of time before entry to the archaeological record, therefore giving earlier date ranges than the rest of the context.) The fact that the Akrotiri samples where 'short-lived' makes an explanation for these early date ranges difficult and, in fact, the older dates from Akrotiri have not yet been satisfactorily explained.
Outliers Due to Possible CO2 Deficiency
Samples contaminated in situ or badly determined by inferior dating laboratories often give rise to outliers (i.e. determinations that lie outside a statistically valid group). Weninger (1990, 218) takes two of the Akrotiri outliers as determined by the Pennsylvania laboratory (samples P-2560 and P-2561) which gave date ranges of 3980 ± 70 BP (2491 ± 105 cal BC) and 3800 ± 50 BP (2235 ± 90 cal BC) respectively and points out that they are centuries too early compared to the rest of the context. It has been shown by Bruns et. al. (1980; see also Michael 1978) that plants found growing too close to Thera have pseudo-ages due to absorbing old CO2 from this volcano. These ages appear to be dependent upon the distance of the samples from the source, therefore giving a measurable decrease the further from the CO2 source that the plant is growing. As well as the concentration of the CO2 source the samples may also be significantly effected by atmospheric circulation. From this it may be argued that samples P-2560 and P-2561 were greatly affected by volcanic CO2 devoid of carbon-14, but this would, by implication, indicate contamination of other samples in the same way (even if this was to a lesser extent). However, there is no gradual tailing off in the dates of the other samples, rather, the others appear as a consistent group. This would be very unlikely if carbon-14 deficiency is the reason for the early dates associated with P-2560 and P-2561 as other samples come from the same event horizon which display no effects that could be due to CO2 emissions. The only reasonable conclusion to be drawn from this is that samples P-1560 and P-1561 were contaminated by an unknown element before they entered the archaeological record.
Is There a Role for Carbon-14 Dating?
For carbon-14 dating to play a more influencial role in the sequential chronology of the Aegean it would be extremely helpful if a greater amount of seriated material was available to help with the process of archaeological wiggle matching (Weninger 1986). That is, determining a multi-phase calibration for an entire regional stratigraphy and simultaneously applying this to a suitable calibration curve (e.g. Stuiver and Becker 1993 decadal curve), thus providing an extremely accurate set of data, "comparable to a fingerprint through time" (Housley et. al. 1999). Nevertheless, estimating calendrical years of archaeological samples inevitably leads to errors in the sample sequence. These may be offset to an extent by applying quantitative archaeological methodology and can be held in check through the use of ceramic seriation, but ultimately, the best samples, "even high-precision dates will only be chronologically useful if the samples derive from secure archaeologically dated contexts" (Shaw 1985, 298).
Although modern determinations are arrived at through the use of high-precision calibration curves after extensive chemical pre-treatment of samples from well defined archaeological contexts, the accurate but ambiguous results obtained mean that radiocarbon dating will always be prone to claims of falibility. Although radiocarbon dating is not a particularly exact method, the evidence it provides does not rule out a late seventeenth-century BC eruption date for Thera
It is true that Popham (1990) has claimed that the carbon-14 determinations from the Egyptian short period occupation site of Amarna (traditionally 1352-1336 BC) support the conventonal chronology, but these are only date ranges and can equally support an earlier date. On the whole, there seem to be too many uncertainties surrounding the use of radiocarbon as a precise dating tool, even when samples are obtained from undisturbed contexts. It has been expressed "that the inherent form of the calibration curve, or rather wiggle, introduces margins of error at 1σ (not to mention 2σ) sufficient to produce a greater degree of imprecision in dates for periods than absolute dates for those periods derived from crosslinks to Egypt" (Warren and Hankey 1989, 127-28). Fortunately, however, science does provide other, more accurate dating techniques.
Greenland ice-cores
The ice-sheets of Greenland can be cored and the contents analysed to reveal environmental evidence stretching back more than 250,000 years (Daansgard et. al. 1993). As annual snowfall has been compacted over this period to form layers of ice, the various ice-cores can be used to provide layered acidity records which can be roughly correlated with known volcanic eruptions. Examination of the GISP-2 ice-cores has revealed volcanic glass (tephra) but chemical analysis of this has proved to be debatable. The method of simply counting back through the ice layers in an attempt to provide calendrical dates is not entirely precise as even stratified tephra can only be used to provide a sequential chronology due to uncertainties concerning past snowfall leading to missing or multiple layers in the ice-cores.
Hammer et. al. (1987) have used the volcanic explosivity index (VEI) to assign a magnitudal value of 6 to the eruption of Thera. This compares to the AD 1883 eruption of Krakatoa level of 6 and the AD 1815 eruption of Tambora level of 7. These eruptions were of huge magnitude and the aerosols expelled by them into the stratosphere eventually reached the Greenland ice-sheets as a result of precipitation.
Of the ice-cores taken from sites in Greenland, all of them reveal high levels of sulphuric acid present around the proposed time of the Thera eruption (Vitaliano et. al. 1990; Soles et. al. 1995; Manning 1998; although cp. Zielinski and Germani 1998a; 1998b). It may be worth mentioning that it was previously thought that a particularly strong acid peak visible in the Camp Century (north Greenland) ice-core dated to around 1390 BC may have been caused by the Minoan eruption (Hammer et. al. 1980), but as this did not correlate with Dye-3 (south Greenland) it was assumed to be an unknown high-latitude eruption. Nevertheless, the sulphate peak at 1644 ± 20 BC from Dye-3 has been linked by Clausen et. al. (1997, 26,713) to one in the more recent GRIP ice-core at 1636 ± 7 BC which has already been associated with the eruption of Thera (Hammer et. al. 1987, 517). The GISP-2 core contains an acidity peak at 1623 ± 36 BC (Zielnski and Germani 1998a) and Camp Century centres on 1598 ± 30 BC (Hammer 1980; 1984).
The similarities in the ice-core date ranges have been noted by Baillie who makes the interesting and probably quite significant point that the error limits associated with each sulphur peak at this time can be correlated with a series of narrow ring events found in northern European dendrochronologies (Baillie 1994; 1996b, 708). This would seem to indicate that we are observing a major climatic event on a global scale. Unfortunately, although volcanic tephra has been found in one GISP-2 ice-layer, there is not yet definitive evidence that this originates from Thera. If Theran tephra could be found in an ice-layer, even though these cannot be precisely dated, the error limits associated with these are sufficiently narrow to be able to eliminate one or other possible eruption candidates (i.e. late seventeenth- or late sixteenth-century BC).
Evidence of Volcanism from the Dye-3 Ice-Core
Analysis of the Dye-3 ice-core established the presence of three high-level acidity peaks in layers that could be roughly associated with the proposed date of the Theran eruption. However, after chemical analysis it was decided that the peaks at 1688 BC and 1428 BC where composed, in the main, of nitric acid. These indicated summer melt rather than emanating from volcanic activity. Only the 1644 BC signal appears to have a high enough sulphuric acid content to be of volcanic origin.
Although Hammer et. al. claim "the Dye-3 area receives amounts of annual precipitation to rule out the possibility of missing annual snow deposits" they fail to reveal how they know this or how indeed it would be possible to notice a missing annual snowfall. Nevertheless they have gone on to suggest 1645 ± 7 BC as the date of the Theran eruption based on the strong sulphur acidity peak showing in the layer believed to represent 1644 BC (Hammer et. al. 1987, 519). However, it has subsequently been shown that the high sulphuric acid peak from the Dye-3 ice-core is too intense to represent Thera, indicating that it must signify another unknown eruption. This view finds support in Warren who states that "the 1645 BC event marked by the ice-core is not connected to the Theran eruption" (Warren 1990b, 33).
More Evidence from GISP-2
Another large acid peak from Greenland is found in the GISP-2 ice-core and provides a date of 1623 ± 36 BC (Zielinski and Germani 1998a). This is definitely caused by volcanic activity as tephra has been uncovered from the stratigraphy which should, in theory, be able to accurately identify the actual volcano involved. However, the tephra has been analysed and does not appear to correlate with any known volcanic eruption from this period. Zielinski and Germani take these results, along with other volcanic and tree-ring evidence, to indicate than another, undocumented, volcano is responsible for "the many climate-proxy signals in the late 1620s BC" and that 1627/28 BC should no longer be held as the definitive age of the Santorini [Theran] eruption" (Zielinski and Germani 1998a,279).
Manning has taken issue with this proposal, claiming that the four sherds of volcanic glass examined by Zielinski and Germani (1998a) from the GISP-2 ice-core do indeed bear a close resemblance to "the most evolved, initial, glass products of the Santorini eruption" (Manning 1998, 1039). In support of Manning, an analysis of Theran glass by Soles et. al. (1995, 387-390) and Vitaliano et. al. (1990) revealed that some, at least, of this glass was indeed compatible with the GISP-2 shards as is the alkiline composition between the GISP-2 shards and those from Thera. It is also apparent that both the GISP-2 and Theran evolved glass contained low readings of MgO and TiO2 which differed from other evolved glass found on Mochlos and the initial Rose Pumice on Thera to later basic glass deposits and bulk tephra samples on Thera. The different glasses also show various degrees of evolution which, although slight, may indicate that Mochlos tephra evolved high within the magma chamber and was therefore the first to erupt and travel the furthest.
The implication of Manning's re-evaluation of the Zielinski and Germani paper is that far from indicating another (historically unknown) volcanic eruption, the GISP-2 ice-core is actually picking up the signal from Thera in the 1623 ± 36 BC layer. Moreover, the fact that the GISP-2 ice-core is incomplete (87m of the core was unsuitable for analysis from a depth of c. 767m making precise multi-core replication impossible (Alley et. al. 1993, 528) introduces an unknown error to the accuracy of the dating. (GISP-2 is not alone in this, Hammer et. al. (1987, 518-519) tell us that the ice-core from Camp Century is also incomplete, particularly during the period 2000-1600 BC.) This additional error factor produces a further element of flexibility to the date of 1623 ± 36 BC. Whenever Clausen et. al. compared the GRIP and GISP-2 ice-cores they found that during the second millennium BC "the uncertainties of the two timescales deviates too much to make a safe comparison" (Clausen et. al. 1997, 26713).
Nevertheless, Zielinski and Germani have restated their case (1998b) by providing a series of plots to compare the chemical composition of the volcanic glass in the GISP-2 1623 ± 36 BC layer. They conclude by reasserting their original view that the evidence points to another eruption to that of Thera on the basis that there is no match in composition or any significant overlaps between the GISP-2 glass and that deriving from Thera. They also disagree with Manning's assertion that the Mochlos and Rose Pumice glass was evolved glass and travelled the furthest. Zielinski and Germani furthermore claim that the initial eruptive sequence need not be the most explosive phase but is more commonly a smaller plinian eruption. This argument finds support from the Nile Delta where archaeological evidence apparently agrees with the view that "the eruptian sequence included a plinian pumice fall deposit" (Stanley and Sheng 1986, 733). However, regardless of the dating of this eruption, and whether or not it represents Thera, Zielinski and Germani are in agreement with Manning that it occurred around 1623 ± 36 BC.
In conclusion, it may be stated that, although there is no agreement as to the actual candidate, the various ice-core records do indicate that a very significant volcanic eruption occurred somewhere in the northern hemisphere towards the end of the seventeenth-century BC. No reason exists, other than tradition why this eruption cannot be Thera. For example, Bruins and van der Pilcht (1996, 213) state that "the Minoan eruption at Santorini produced the largest volcanic dust cloud over the eastern Mediterranean in the second millennium BC", while Sullivan points out that Theran tephra found at Golcuk supports the theory of a major, global, climatic event due to the estimated 13km3 volume of dense rock ejected (Sullivan 1998; see also Polinger and Ritner 1996).
Dendrochronology
This discipline takes the relatively straighforward process of annual tree-ring growth to produce precise, exact, calendrical determinations. Each year trees grow an additional ring with the size of each depending, to a large extent, on how warm or cold the temperature was. The ring patterns can then be matched against others from the same tree species to produce environmental records of regional warming and cooling events. These patterns can then be fitted onto a master dendrochronology and correlated with similar records in widely different geographical locations to show the limits of an environmental event. The enviromental information provided by dendrochronology can therefore be expressed in terms of historical time.
The importance of dendrochronology and its role as a precise dating tool has increasingly been recognised by archaeologists. This technique provides a level of accuracy unsurpassed by other scientific dating methods. Warren has summed up the situation well in stating that "calendrical probabilities are just that, degrees of probablity . . . I am sure however that the real future for Aegean absolute dating lies not with radiocarbon . . . but with dendrochronology" (Warren 1996, 283-284). A major advantage with this technique are the environmental correlations which may be found between separate regions. "We can look at what was happening to trees in different areas at exactly the same time" (Baillie 1996a, 291).
Using Dendrochronology to Link Various Scientific Dating Techniques
The evidence produced by the tree-rings is not in doubt, a clearly indicated climatic event occurred in 1628 BC which shows up in Irish bog oaks as a narrowest-ring event (Baillie and Munro 1988) and in the 1627 BC tree-rings on Californian bristlecone pines as frost damage (LaMarche and Hirschboek 1984). Baillie has further shown the likelihood of these tree-ring signals being associated with the GISP-2 and Dye-3 Greenland ice-core dates of 1623 ± 36 BC and 1645 ± 20 BC respectively which also indicate major volcanic activity at this time. Significantly, the estimated error limits associated with both these ice-core determinations overlap the precise dendrochronologically derived date of 1628 BC (Kuniholm 1990). Furthermore, Johnson et. al. (1992) have been able to show that there were only three significant acid layers present in the GRIP ice-core (2039 ± 5 BP, 3636 ± 7 BP and 4040 ± 10 BP; with BP calculated from AD 1990). The 3636 ± 7 BP date would appear to correlate with the 1627 BC tree-ring evidence therefore providing similar results between dendrochronology and the ice-cores seemingly indicating that they are both picking up the signal of the same event. Although additional tree-ring evidence comes from England, Germany and Turkey (Baillie 1990), it should be noted that dendrochronology only provides confirmation of a climatic event, it cannot confirm volcanic causation, much less which actual volcano. That is to say that the narrow or frost damaged tree-rings indicate a climatic downturn, but that this may not necessarily be caused by the emission of volcanic aerosols into the atmosphere. It is extremely frustrating that the only exact scientific dating method is unable to answer this question.
However, dendrochronology can supply proxy evidence for Greenland ice-cores showing high sulphate levels around the seventeenth- and sixteenth-centuries BC. Baillie and Munro (1988) were able to demonstrate from Northern Irish bog oaks that there appears to be a correlation between some clusters of narrowest ring events and volcanic eruptions. It was seen that when Northern Irish oaks were arranged into a crude narrowest ring index and the product of the total number of narrowest rings and the number of sites computed for a ten year period, that the three highest values (1153 BC, 3199 BC and 4377 BC) agreed with three (1100 ± 50 BC, 3250 ± 80 BC and 4400 ± 100 BC) of the six main acidity peaks from the Camp Century ice-core (see Hammer et. al. 1980). Baillie and Munro also make an important point regarding the slight variations in dating between the Northern Ireland oak narrowest ring event and the Greenland ice-cores. They state that in such cases "the tree-rings must take precedence because every tree-ring containing frost damage or narrow rings is absolutely dated (Baillie and Munro 1988, 346). Kuniholm's work on Anatolian tree-rings has produced a continuous 1503 year chronology from 2233 BC to 731 BC (+76/-22 years) which he believes provides evidence for the volcanic eruption of Thera between 1641+45/-14 BC as wiggle matched (i.e. multiple determinations simultaneously applied onto the calibraton curve) and more specifically 1628 BC (Kuniholm 1996, 333; see also Kuniholm et. al. 1996).
Doubts About Climatic Assumptions - and their Weaknesses
Not all agree with the claims made by dendrochronologists that the Minoan eruption must have had a global impact causing noticable climatic variations which show up in the tree-rings. While Manning (1990, 35) makes the valid point that not all eruptions leave their mark on the ice-cores or tree-rings, Pyle (1990) goes further claming that no reason exists for the assumption that Thera should furnish the most extreme climatic global effects for the perod between the eighteenth- and forteenth-centuries BC. This seems a strange claim to make in view of the substantial evidence from both ice-cores and tree-rings which indicate no other known candidate of this magnitude (i.e. VEI 6, Hammer et. al. 1987) at this time (see above and esp. Baillie 1996b). Although Pyle goes on to estimate that there may possibly be up to thirty eruptions in this time range of the magnitude required to expel aerosols into the stratosphere (therefore causng frost damage and narrow rings in trees and high sulphuric acidity peaks in ice-cores), he ignores the fact that the signals of these do not appear in either the Greenland ice-cores or the various world-wide dendrochronologies. In short, Pyle's possible thirty eruptions is no more than an uninformed guess whereas the wide distribution of Theran tephra throughout the region certainly suggests that this eruption could have been one of, if not the most violent in post-glacial times.
Conclusion to the Scientific Dating Techniques
Although radiocarbon determinations prove inconclusive, it is clear from the evidence provided by ice-cores and tree-rings that a major climatic event occurred towards the end of the seventeenth-century BC. It is tempting to see in this the volcanic eruption of Thera, especially as no other possible eruption has been forwarded as a realistic alternative candidate. As scientific dating procedures appear to indicate a much earler eruption date for this volcano than that traditionally assigned to it by Egyptian chronology, it raises an extremely important question concerning the start of the Aegean Late Bronze Age.
As there is clearly a close approximation between the main scientific dating methods (Baillie 1996b), it is these determinations which shall form a working hypothesis for this critical analysis of ancient chronology. An eruption date of 1628 BC for Thera will be taken as the basis on which conventional Egyptian chronology shall be subjected to scrutiny. Science provides us with empirical fact while the traditional Egyptian dating system is based on biblical and astronomical data and the interpretation of archaeological discoveries. It would therefore seem advisable - if the eruption date of Thera is to be pushed back in historical time by around a century - to methodically examine the factors that have been combined to produce the conventional Egyptian chronology. This will lead us into other Levantine regions such as Assyria, Babylonia and Hatti. These areas were in contact with Egypt throughout the second millennium BC and have left us extensive archival material from which it may be possible, not only to correlate these regions with Egypt, but also to provide free-standing chronologies for these regions in their own right.