The Olive Branch and the Ten Experts – How the Thera Eruption Is Dated, and Why It Is Taking So Long

The Hook

Imagine three systems logging independently of one another. Each has its own time axis, each axis is internally consistent, and none of the three knows the others' clock. You are asked to reconstruct the actual order in which events happened.

The obvious way out: find an event that appears in all three logs. An outage, a deploy, a network segment that went away for everyone at the same moment. If you can locate that one event on each time axis, you have a common nail on which to hang all three.

The archaeology of the eastern Mediterranean has exactly this problem – and exactly this nail. Pearson and colleagues framed the stakes in Science Advances in 2018: the Thera eruption offers a critically important marker horizon for synchronizing the archaeological chronologies of the Aegean, Egypt and the Near East, and for anchoring paleoenvironmental records from ice cores, speleothems and lake sediments.

Except that the nail does not hold. Two methodologically different approaches – radiocarbon dating and the archaeological-historical linkage through Egypt – have for decades delivered answers that were far apart. Sturt Manning writes in PLOS ONE in 2022 that archaeologists long placed the eruption around 1500 BCE and therefore in the Egyptian New Kingdom, whose beginning he gives as roughly 1565 to 1540 BCE, whereas radiocarbon pointed to dates at least 50 to 100 years earlier, that is, in the preceding Second Intermediate Period. The current discrepancy, Manning writes in his introduction, has led to rival and largely incompatible "high" and "low" chronologies and different histories.

Fifty to a hundred years is not a nuance in this period. It decides which political world the eruption falls into.

This article is less about volcanoes than about measurement chains: how many links sit between a physical measurement and a calendar year, which of those links have proved particularly fragile in this case, and what that teaches about handling derived numbers of your own. It is also, incidentally, one of the finest error-correction stories in recent science – because nearly every attack on a result here is published, answered, and available to read.

The Measurement Chain Nobody Holds in Their Head

When people speak casually of "radiocarbon dating", it sounds like a measurement: sample in, calendar year out. In fact it is a chain. The breakdown below is textbook knowledge that I did not verify against a primary source in this run; the evidence in this article concerns the steps that follow.

1. Measured is the amount of radioactive carbon-14 in a sample – a physical value with measurement uncertainty.
2. From that a "radiocarbon age" is computed, a number in conventional radiocarbon years. That is not yet a calendar date.
3. Translated is this radiocarbon age via a calibration curve. That curve is itself a measurement result, obtained from material of known calendar age – Pearson and colleagues write explicitly in 2018: "using calendar-dated tree rings".
4. Interpreted is the result, finally, in the light of further assumptions: does the sample really belong to the event? How many years lie between the measured material and the event? Which stratigraphic sequence counts as secure?

The calibration curve in step 3 is versioned. Friedrich and colleagues name IntCal04 explicitly in 2006 as the curve they match against; the University of Arizona releases speak of IntCal13 in 2018; Bruins and van der Plicht assign Manning's range to the curve IntCal20 in 2025. Three curve versions in roughly twenty years, and each of them can translate the same raw measurements into different calendar years.

This case is instructive because every single link in the chain has at some point been publicly attacked. Let us start with the most spectacular link.

2006: A Branch Buried Alive

Walter L. Friedrich, Bernd Kromer, Michael Friedrich, Jan Heinemeier, Tom Pfeiffer and Sahra Talamo published a paper spanning a single journal page in Science on 28 April 2006: "Santorini Eruption Radiocarbon Dated to 1627–1600 B.C.", volume 312, issue 5773, page 548, in the short-format section Brevia.

The methodological appeal lies in the object. The authors report a branch from an olive tree that was buried alive in tephra on Santorini, with branches of the crown partly preserved in life position. As support for it having been alive when buried, they cite the horizontal position of seven molds of branches in the pumice one to three metres above its base, as well as remnants of olive leaves and twigs covered by the pumice.

Why is this so attractive? A tree buried while alive stops growing at the moment the event occurs. Its outermost annual ring is then – if the assumption holds – immediately adjacent to the event. The paper accordingly does not distinguish between the ring date and the eruption date; the equation is its premise, not its result.

The procedure: the authors used x-ray tomography to identify 72 rings in a section of the branch with preserved bark. They divided this section into four consecutive groups of rings and dated each separately; the figure caption gives the groups as 13, 24, 22 and 13 rings, 72 in total. Four measurements at known spacing yield not four independent point estimates but a sequence which, in the authors' own phrasing, can be "wiggle-matched" against the calibration curve IntCal04. Using the calibration program OxCal they determined the calibrated age range of the outermost ring as 1621 to 1605 BC at 68 percent confidence and 1627 to 1600 BC at 95 percent.

For comparison they state that until then it had not been possible to date the eruption more precisely than the range 1663 to 1599 BC at two standard deviations. And they draw a hard line: a date around 1520 BC or later, as assumed by some archaeologists working with Egyptian contexts, is not consistent with their result even within three standard deviations, that is, 99.7 percent confidence. That exclusion claim, however, is computed against IntCal04 – against precisely the reference whose offset was reported twelve years later.

The authors also name a weakness themselves: olive trees, they write, form irregular, barely visible rings. The number 72 is therefore not a counting result but an interpretive one. They test it: even taking into account an uncertainty of 50 percent in the ring count, potentially caused by growth irregularities of olive, the limits increase by only a decade. And they argue against an interfering effect: they excluded a significant local offset of the radiocarbon ages by volcanic carbon dioxide, because it would then be impossible to match their radiocarbon sequence anywhere to the shape of the calibration curve – an argument that pits the shape of the data against an interference rather than its absolute position.

It was precisely at this self-test that the criticism took aim.

2014: Ten Experts, One Branch, No Agreement

In 2014 Antiquity published a debate under the title "Bronze Age catastrophe and modern controversy: dating the Santorini eruption" (volume 88, pages 267–291). The main contribution, "The olive-branch dating of the Santorini eruption" (pages 267–271), was written by Paolo Cherubini, Turi Humbel, Hans Beeckman, Holger Gärtner, David Mannes, Charlotte Pearson, Werner Schoch, Roberto Tognetti and Simcha Lev-Yadun.

Their attack is empirical and uncomfortably direct. They report a blind test involving several tree-ring laboratories, in which tree rings from olive trees currently growing on Santorini were to be dated. That test, they write, clearly showed that measurements of tree rings in olive wood from Santorini are highly unreliable, owing to intra-annual wood density fluctuations, to variability in tree-ring boundary structure, and to a restriction of cambial activity to shifting sectors of the circumference – which causes the tree-ring sequences along radii of the same cross sections to differ.

The numbers are the argument. Ten tree-ring experts took part in the study. The average number of tree rings counted by the ten experts showed maximal deviations between 24.5 percent and 56.3 percent from the median. From this the authors derive a calculation: a difference of 44 percent – the average deviation in the olive tree-ring measurements by the ten experts – applied to the 72 putative tree rings described by Friedrich and colleagues would result in a range of 40 to 104 years, rather than the 72 proposed. According to Cherubini and colleagues, Friedrich and colleagues had estimated a maximal possible error of plus/minus 3 years for each of the four segments of the olive branch examined, giving a total of plus/minus 12 years.

Their conclusion: the dating of the Thera eruption based on the putative tree-ring sequence from a single olive tree must be considered with great caution. And further: accordingly, caution should be applied to the dating offered by Friedrich and colleagues, and their proposal cannot be used to discount the date range of 1525 to 1490 BC proposed for the eruption from numerous other radiocarbon studies.

Friedrich and colleagues replied in the same debate. Their response carries the title "The olive branch chronology stands irrespective of tree-ring counting". Their argument: a sequence model without ring counts still gives a 95.4 percent minimum of 1609 cal BC for the outermost section.

That is a notable defence, because it removes the attacked link rather than repairing it: if the ring count is unreliable, compute without it and see what remains of the result. I am of the opinion that this is the methodologically clean answer to such criticism – irrespective of who ends up holding the correct date.

One more thing is worth noting: Charlotte Pearson, a co-author of this criticism, is the lead author of the two papers that reorganized the calibration side in the years that followed.

2018: The Modern Olive Tree as a Control Experiment

The most elegant way to test an archaeological assumption is to try it on a living object whose age you know. That is exactly what Yael Ehrlich, Lior Regev and Elisabetta Boaretto did.

Their paper "Radiocarbon analysis of modern olive wood raises doubts concerning a crucial piece of evidence in dating the Santorini eruption" appeared on 9 August 2018 in Scientific Reports, volume 8, article number 11841.

The method exploits a peculiarity of the twentieth century. By their own account, the authors analyzed radiocarbon concentrations in cross-sections from a modern olive tree trunk as well as from a living branch, and obtained near-annual resolution dates using the radiocarbon "bomb peak". That this spike goes back to atmospheric nuclear weapons testing is textbook knowledge which I did not verify against a primary source in this run; what is documented is that it enables the near-annual resolution.

The result, in the abstract: in both cases they show that radiocarbon dates of the last formed wood along the circumference are not chronologically homogenous, and can differ by up to a few decades. The outermost wood layer therefore does not necessarily represent the date of the last year of growth. These findings challenge the interpretation of the results obtained from dating the olive branch from the Santorini eruption, as it could predate the eruption by a few decades.

The individual figures are drastic. For one of the two trees, only 3 of 11 wood samples taken from just beneath the bark gave a date of 2009 or later; all the others were older, and the discrepancy can be as much as 40 years. For the second tree the outermost samples dated in the range 2002 to 2009, while five samples below the bark produced older dates – the oldest around 1955.

This does not strike the ring count but a still more fundamental assumption: that "outer" means "young". In a tree whose cambial activity shifts between sectors, the outermost layer at one point of the circumference is not necessarily the same age as the outermost layer at another.

In the same year, 2018, the second attack arrived – aimed at step 3.

2018: It Was Not the Measurement That Was Off, but the Yardstick

Charlotte L. Pearson, Peter W. Brewer, David M. Brown, Timothy J. Heaton, Gregory W. L. Hodgins, A. J. Timothy Jull, Todd Lange and Matthew W. Salzer published, in Science Advances in 2018, volume 4, issue 8, article number eaar8241, the paper "Annual radiocarbon record indicates 16th century BCE date for the Thera eruption". (I take the spelled-out first names from the release by the University of Arizona's tree-ring laboratory; the bibliographic query returns only abbreviated forms.)

Their point of attack was not the sample material from Thera but the calibration itself. By their own abstract, they used calendar-dated tree rings to create an annual resolution radiocarbon time series for 1700 to 1500 BCE in order to validate, improve, or more clearly define the limitations for radiocarbon calibration of materials from key eruption contexts.

The core sentence reads, in the original: "Results show an offset from the international radiocarbon calibration curve, which indicates a shift in the calibrated age range for Thera toward the 16th century BCE."

The abstract names no numeric value for the offset, no calendar window in which it operates, and no year range for the eruption. I was unable to open the full text in this run; I therefore make no statement about the magnitude or sign of the offset.

The University of Arizona releases name bristlecone pine wood and Irish oak wood from County Kildare as the material, and a total of 285 samples – 200 annual samples from bristlecone pines for 1700 to 1500 BCE and 85 Irish oak samples from the same period. The two releases differ on one point: the tree-ring laboratory page names only the White Mountains of California, while the university news page says "California and Nevada". These quantity and location figures come from press material, not from the paper.

Pearson and colleagues also state the conflict in its then-current form in the abstract: the finding sheds new light on the long-running debate focused on a discrepancy between radiocarbon dating evidence for Thera – late 17th to early 16th century BCE – and archaeological dating evidence – mid 16th to early 15th century BCE.

The progress in this paper is not only chronological but also epistemic: part of the difference sat not in a clash of methods but in the reference curve used by the radiocarbon side alone.

A word of candour about the explanation at this point. That calibration curves were historically obtained in many cases from wood blocks aggregated over several years, and that such smoothing can produce systematic shifts, is the obvious explanation for an offset of this kind – but it is textbook and plausibility knowledge that I was unable to verify against any primary source in this run; the relevant paper in Radiocarbon on the calibration of multiple tree-ring blocks was not accessible. I therefore state the mechanism explicitly as an unverified conjecture, not as a finding.

2021: Partial Rehabilitation – With a Catch

The same group that had sown doubt in 2018 supplied the counter-test in 2021. Yael Ehrlich, Lior Regev and Elisabetta Boaretto published, on 12 January 2021 in Scientific Reports, volume 11, article number 704, the paper "Discovery of annual growth in a modern olive branch based on carbon isotopes and implications for the Bronze Age volcanic eruption of Santorini".

The method is a pretty one. Instead of counting rings, they looked for the annual rhythm in the carbon isotope ratio: in their own phrasing, the delta-13C values of the contiguous cellulose samples from the olive wood segment revealed a recurring cyclic pattern. They divided a six-centimetre segment covering the years 1957 to 1972 into 96 contiguous samples.

Their result, verbatim: this shows that indeed, olive trees do produce new wood annually and that this wood can be identified by the delta-13C pattern, making this 15-year segment the first sequence of verified annual growth rings in olive wood.

That is a rehabilitation of the underlying assumption – with an explicit catch. The abstract states: although these results support the validity of the Santorini branch date, they observed some chronological anomalies in modern olive and simulated possible date range scenarios of the volcanic eruption of Santorini given these observed phenomena. And the closing sentence of the abstract: the results offer a way to reconcile this long-standing debate towards a mid-sixteenth century BCE date.

How large the catch is becomes clear from an individual observation the authors record verbatim: between the ring dated to 1971 and the outermost wood dated to 2005–2009 there are fewer than ten possible rings visible. Thus, just within the last 1.5 centimetres of the section, there is a discrepancy of about 30 years between the number of rings identified by micro-CT and the years measured by radiocarbon.

Thirty missing years in one and a half centimetres – and precisely with the method, micro-CT, by which the 72 rings were determined in 2006.

2020: How to Write Down a Weak Finding Correctly

A brief sideways glance, because it contains a methodological model.

Charlotte L. Pearson, Matthew W. Salzer, Lukas Wacker, Peter W. Brewer, Adam Sookdeo and Peter Kuniholm published, in the Proceedings of the National Academy of Sciences in 2020, volume 117, pages 8410 to 8415, the paper "Securing timelines in the ancient Mediterranean using multiproxy annual tree-ring data".

They fix a "floating", that is, not yet absolutely dated, East Mediterranean tree-ring chronology more closely in time using annually resolved patterns of carbon-14 modulated by cosmic radiation between 1700 and 1480 BC. This placement is then tested using an anticorrelation between calendar-dated tree-ring growth responses to climatically effective volcanism in North American bristlecone pine and the Mediterranean trees – two different lines of evidence, the second of which tests the first.

Then comes the actual finding: examination of the newly dated Mediterranean tree-ring sequence between 1630 and 1500 BC using X-ray fluorescence revealed an unusual calcium anomaly around 1560 BC. The authors write that, while requiring further replication and analysis, this anomaly merits exploration as a potential marker for the eruption of Thera.

That is not a dating of the eruption. It is a candidate for a marker, explicitly flagged as requiring replication. Turning a sentence of that construction into a headline reading "Researchers date Thera to 1560 BC" is a widespread error type in reading primary literature – and the reason this article so often states what a source does not say.

2022 and 2024: Bayes and the Stratigraphy

Sturt W. Manning, Cornell Tree-Ring Laboratory, published on 20 September 2022 in PLOS ONE, volume 17, issue 9, article number e0274835: "Second Intermediate Period date for the Thera (Santorini) eruption and historical implications".

Manning tackles two things. First, the volcanic carbon dioxide objection: volcanic CO2 was often blamed for the discrepancy, but a comparison of radiocarbon dates directly associated with the eruption from contemporary Aegean contexts – both on and remote from Thera – can now remove this caveat. The abstract does not state the outcome of that comparison explicitly; the underlying logic, that only matching ages on and off the island can rule out an island effect, is my own addition.

Second, the analysis: using Bayesian analysis, a revised and substantially refined date range can be determined, through the integration of the large radiocarbon dataset relevant to the eruption with the local stratigraphic sequence on Thera immediately prior to the eruption, and in conjunction with the wider stratigraphically-defined Aegean sequence from before to after the eruption. The result, in the original: "~1606–1589 BCE (68.3% probability), ~1609–1560 BCE (95.4% probability)". Manning himself calls this a "robust high-resolution dating" and explicitly claims substantial progress.

Century arithmetic is needed here, because it is often done wrong. The 17th century BCE runs from 1700 to 1601 BCE, the 16th from 1600 to 1501. Manning's 68.3 percent range of 1606 to 1589 therefore straddles the turn of the century: six years in the 17th, twelve years in the 16th century. His 95.4 percent range extends to 1560 and therefore into the middle of the 16th century.

Can this be reconciled arithmetically with Pearson 2018? No – and this is a point to keep clean. Pearson and colleagues name no year range at all in their abstract, only a direction. Two ranges cannot be made to overlap when one of them does not exist. What is documented is only this: Manning's range reaches into the same century toward which Pearson's offset shifts. No contradiction follows from that, but no confirmation either.

According to Bruins and van der Plicht, Manning followed up in 2024: their introduction cites, for Manning's 2024 analysis, ranges of 1612 to 1602 or 1613 to 1602 BCE at one standard deviation and 1616 to 1589 or 1618 to 1584 BCE at two standard deviations, with a maximum around 1608 BCE. I did not open the 2024 paper itself in this run; I therefore quote these figures second-hand. If reported correctly, they show a shift back relative to 2022, toward the end of the 17th century.

Historically, Manning in 2022 places the eruption in the Second Intermediate Period with its Canaanite-Levantine dominated world-system, and sees in this an importantly altered cultural and historical context for the New Palace Period on Crete and the contemporary Shaft Grave era in southern Greece. In addition, the revised dating, together with a current absence of southern Aegean chronological data placed soon afterwards, highlights a period of likely devastating regional eruption impact in the earlier-mid 16th century BCE southern Aegean.

2025: The Other Side of the Equation

And then something happens that one rarely sees in dating debates: the dispute is resolved not by one side moving, but by measuring the other number.

Hendrik J. Bruins (Ben-Gurion University of the Negev) and Johannes van der Plicht (University of Groningen) published on 10 September 2025 in PLOS ONE, volume 20, issue 9, article number e0330702: "The Minoan Thera eruption predates Pharaoh Ahmose: Radiocarbon dating of Egyptian 17th to early 18th Dynasty museum objects".

The approach: instead of dating Thera once again, they date the Egyptian side. By their own abstract, they investigated museum objects from the 17th and early 18th Dynasty, at the transition from the Second Intermediate Period to the New Kingdom, a period hardly studied with radiocarbon dating. Their research facilitated the first-ever direct radiocarbon time comparison between this dynastic transition period and the Minoan Thera eruption.

The objects named in the abstract: a mudbrick from the Ahmose Temple at Abydos (British Museum), a linen burial cloth associated with Satdjehuty (British Museum), and wooden stick shabtis from Thebes (Petrie Museum).

Methodologically interesting is what they cannot do. Since these items cannot be arranged in a stratigraphic sequence, Bayesian analysis could not be used; they adopted an alternative strategy within radiocarbon time space. Concretely: they compare their uncalibrated dates for the objects with a robust series of uncalibrated radiocarbon dates for the Thera eruption – and find that the two data sets have a different time signature.

That is a methodological move whose logic deserves appreciation: comparing two data sets on the same, uncalibrated axis bypasses the calibration link of the measurement chain entirely. Whatever distorts the curve distorts both sides equally – and the comparison remains valid even when the absolute calendar year stays open.

The result, in the abstract: the Minoan eruption is older than the reign of Nebpehtire Ahmose, the first king of the 18th Dynasty, who reunited Upper and Lower Egypt. Their calibrated results support a low chronology for his reign and the beginning of the New Kingdom. Previous radiocarbon dates of king Senusert III, by contrast, support a high chronology for the Middle Kingdom. Therefore the Second Intermediate Period, sandwiched in between these united Egyptian Kingdoms, embodies a significant time interval.

The point is structural. For decades the question read: does Thera move later, or does Egypt stay early? This paper answers: Egypt moves later. The Egyptian anchor against which the Aegean chronology was hung shifts – and the Second Intermediate Period, which had been treated as a short interregnum, grows longer.

Restraint is warranted here too: the abstract names no numeric value with a confidence level for the start of Ahmose's reign. The press releases on this paper do give numbers – roughly 1539 instead of roughly 1580 BCE – but without a confidence statement and not from the abstract. I therefore do not adopt them as a finding.

The Ice Core That Was a Different Volcano

A brief side strand, because it contains the single most instructive case.

A signal in the Greenland ice dated to about 1645 BC was for a time taken as the trace of the Thera eruption. N. Pearce, J. Westgate, Shari J. Preece, W. Eastwood and W. Perkins published in 2004 a paper whose title carries the punchline: "Identification of Aniakchak (Alaska) tephra in Greenland ice core challenges the 1645 BC date for Minoan eruption of Santorini". The journal is recorded in the bibliographic query only as "Geochemistry", volume 5; the abstract and full text were not accessible in this run, which is why I cite this source at title level only: the authors identify tephra from the Aniakchak volcano in Alaska in the Greenland ice core, and thereby challenge the attribution of the 1645 BC date to the Minoan eruption of Santorini.

This is the most uncomfortable way to lose a piece of evidence: it was not the counting of the ice layers that was called into question, but the attribution of the signal to an event. In the measurement chain above, that is step 4 – and step 4 is the one that tighter statistics do not replace.

A Framework: How Solid Is What?

What Remains Open

There is no single calendar year that emerges as secure from the papers opened here.

What there is, is a shift in what is being disputed. Before 2006, the range 1663 to 1599 BC (2 sigma) was the best available, according to Friedrich and colleagues. In 2006 came the narrower window of 1627 to 1600 BC (95%). Manning's 95.4 percent window of 2022 runs from 1609 to 1560 – at 49 years, therefore wider again than Friedrich's 27 years. The progress lies not in the width but in the stratigraphic grounding and in the relocation. And the most recent step, in 2025, shifts not Thera but the Egyptian reference point.

Not answered in the sources checked here: the magnitude of the offset found by Pearson and colleagues and the calendar window over which it operates – that is not in the abstract, and I could not open the full text. Whether the calcium anomaly around 1560 BC is replicable – open, according to its own authors. How Manning's 2024 analysis argues in detail – I have it only second-hand. And how the findings of Ehrlich and colleagues relate to Manning's range when both are taken together – none of the papers opened here attempts that explicitly, and I do not attempt it here either.

I am of the opinion that this is not a failure but the normal state of a well-functioning discipline. What is most remarkable is the form: nearly every support for this date has been attacked – the ring count, the outer-means-young assumption, the calibration curve, the ice core attribution, the Egyptian anchor – and every one of those attacks is published, numbered and answered. That is error correction in production, with a complete audit log.

The Central Takeaway

The transferable lesson lies not in the history but in the measurement chain.

Test the assumption that makes your result elegant – it is usually the weakest one. The olive branch was attractive because it was buried alive; everything therefore hung on "buried alive", "72 rings" and "outer means young". Precisely those three points were attacked, two of them by control experiments on modern trees. What makes a result especially convincing deserves the hardest scrutiny, not the mildest. In practice: the path on which your system performs conspicuously well is the path you should be writing the test for.

A derived value without the version of the reference it was derived from is incomplete. Friedrich and colleagues write IntCal04 into their paper; that is why, twenty years later, it remains traceable which reference their result was computed against and how it would be revised against a newer curve. If you store amounts, risk classes, geolocations or compliance assessments in a data set, store the reference version and the time of derivation alongside the value. Otherwise you will not be able to tell later whether the world changed or only your lookup table. That such a reference can have three versions in twenty years is exactly what this case shows, with IntCal04, IntCal13 and IntCal20.

Compare before normalizing, wherever you can. The strongest methodological move in this story is the one by Bruins and van der Plicht in 2025: compare two data sets on the uncalibrated axis instead of sending both through the error-prone translation first. If you want to hold two measurement series against each other, compare them as early in the processing chain as possible – before both have passed through conversions, roundings and lookup tables. A shared distortion cancels out in a comparison; two different post-processings do not.

And record uncertainty as an interval with a level, not as a number. "1627–1600 BC at 95 percent" is a checkable statement; "around 1600 BC" is not. Manning's 68.3 percent range straddles the turn of the century, his 95.4 percent range reaches into the middle of the 16th century: the same data, two noticeably different histories, depending on the confidence level chosen. Cutting the level cuts away the information that makes the value interpretable in the first place.

Reflection Question

Which number in a system you are responsible for is treated as a fact but is actually derived from a reference whose version is stored nowhere – and which assumption in that system is so elegant that nobody has ever tested it?

Fact Check

Automated AI verification against abstracts, full texts and metadata of the primary sources; two papers accessible only bibliographically or second-hand; no human expert review.

From the first draft, 129 checkable claims were extracted: 41 hard facts, 2 quotations, 28 source attributions and 58 synthesis claims. The check was carried out by a separate reviewer pass that received only the draft and the source dossier, with the brief to find errors. Result: 62 claims confirmed, 13 false, 29 to be weakened, 25 not checkable. Both statements presented as "verbatim" quotations were defective. A targeted search for counter-evidence followed; it added five papers that substantially changed the article's content.

The three most important corrections:

1. The olive branch was presented in the draft as largely unchallenged. The counter-evidence search turned up three published attacks on precisely that piece of evidence – the blind test by Cherubini et al. 2014 with deviations of 24.5 to 56.3 percent among ten experts, the control experiment by Ehrlich et al. 2018 on modern olive trees, and the partial rehabilitation by Ehrlich et al. 2021 with its thirty-year catch. The article has been extended by these three sections; the olive branch is now presented as a disputed, not a load-bearing, piece of evidence.
2. The century arithmetic was wrong. The draft claimed that Manning's 68.3 percent range of 1606–1589 BCE lay "entirely at the end of the 17th century". The 17th century BCE ends in 1601; twelve of the range's eighteen years lie in the 16th century. Two sections whose point rested on this error have been rewritten.
3. The central takeaway rested on a false premise. The draft concluded that part of the difference had sat "in the jointly used reference data", which "both sides computed against indirectly". The archaeological-historical side does not compute against IntCal; what is documented is only an offset in the reference used by the radiocarbon side. The takeaway has been reformulated accordingly and extended by the lesson on comparing before normalizing, which is documented by Bruins & van der Plicht 2025.

Further corrections concerned, among other things: two German translations of English originals presented as "verbatim in the abstract" (now original plus flagged translation); the claim that the case had been "decided" at the calibration curve (deleted, since no source claims a decided case); the presentation of 2018 as "epistemic, not chronological" (the paper's title and abstract are explicitly chronological); the claim of a monotone narrowing of the date windows (1663–1599, then 1627–1600, then 1609–1560 correspond to 64, 27 and 49 years – the last step widens); Manning as a witness for continuing incompatibility (he is describing the situation prior to his work and claims progress); "Kolleginnen", a female-plural form, for an author group of one woman and seven men; "roughly a hundred years" instead of the sourced "at least 50 to 100 years"; the invented expansion of the abbreviation LTRR as "Long Tree-Ring Laboratory"; invented spelled-out first names for Pearce et al. 2004; an unsourced issue number for Pearson et al. 2020; and four unflagged textbook passages, now marked as such or deleted – among them the multi-year wood block mechanism, which supplies the explanation for the 2018 finding and is now explicitly flagged as an unverified conjecture.

Also deleted were the burial of Akrotiri and the distribution of tephra "across half the Mediterranean" (in none of the sources opened), the naming of the three competing time axes as king lists, ceramic phases and stratigraphic sequences (not documented), and the description of the Greenland 1645 signal as a "sulphate layer" (the title of Pearce et al. speaks only of a date).

Second-hand, and therefore not primary-verified, are: the figures for Manning's 2024 analysis (quoted from the introduction of Bruins & van der Plicht 2025), the details of sample material and sample counts for Pearson et al. 2018 (from University of Arizona releases, which contradict each other on one point), and the substantive claim of Pearce et al. 2004 (title level only).

Sources and Further Reading

1. Friedrich, W. L., Kromer, B., Friedrich, M., Heinemeier, J., Pfeiffer, T., & Talamo, S. (2006). Santorini Eruption Radiocarbon Dated to 1627–1600 B.C. Science, 312(5773), 548. https://doi.org/10.1126/science.1125087 — journal page fully opened as a PDF in this run; Supporting Online Material not accessible.
2. Cherubini, P., Humbel, T., Beeckman, H., Gärtner, H., Mannes, D., Pearson, C., Schoch, W., Tognetti, R., & Lev-Yadun, S. (2014). The olive-branch dating of the Santorini eruption. Antiquity, 88, 267–271 (within the debate "Bronze Age catastrophe and modern controversy: dating the Santorini eruption", pp. 267–291, with a reply by Friedrich et al., a comment by Bietak and a closing reply by Cherubini & Lev-Yadun). — debate PDF opened in this run; issue number not captured, closing reply not read verbatim.
3. Ehrlich, Y., Regev, L., & Boaretto, E. (2018). Radiocarbon analysis of modern olive wood raises doubts concerning a crucial piece of evidence in dating the Santorini eruption. Scientific Reports, 8, 11841. https://doi.org/10.1038/s41598-018-29392-9 — abstract and parts of the full text opened in this run; discussion not read verbatim.
4. Pearson, C. L., Brewer, P. W., Brown, D. M., Heaton, T. J., Hodgins, G. W. L., Jull, A. J. T., Lange, T., & Salzer, M. W. (2018). Annual radiocarbon record indicates 16th century BCE date for the Thera eruption. Science Advances, 4(8), eaar8241. https://doi.org/10.1126/sciadv.aar8241 — abstract fully opened in this run; full text not accessible.
5. Pearson, C. L., Salzer, M. W., Wacker, L., Brewer, P. W., Sookdeo, A., & Kuniholm, P. (2020). Securing timelines in the ancient Mediterranean using multiproxy annual tree-ring data. Proceedings of the National Academy of Sciences, 117, 8410–8415. https://doi.org/10.1073/pnas.1917445117 — abstract fully opened in this run.
6. Ehrlich, Y., Regev, L., & Boaretto, E. (2021). Discovery of annual growth in a modern olive branch based on carbon isotopes and implications for the Bronze Age volcanic eruption of Santorini. Scientific Reports, 11, 704. https://doi.org/10.1038/s41598-020-79024-4 — abstract and parts of the full text opened in this run; published 12 January 2021, while the DOI carries the year 2020.
7. Manning, S. W. (2022). Second Intermediate Period date for the Thera (Santorini) eruption and historical implications. PLOS ONE, 17(9), e0274835. https://doi.org/10.1371/journal.pone.0274835 — abstract and parts of the full text opened in this run.
8. Bruins, H. J., & van der Plicht, J. (2025). The Minoan Thera eruption predates Pharaoh Ahmose: Radiocarbon dating of Egyptian 17th to early 18th Dynasty museum objects. PLOS ONE, 20(9), e0330702. https://doi.org/10.1371/journal.pone.0330702 — abstract and introduction opened in this run; published 10 September 2025.
9. Pearce, N., Westgate, J., Preece, S. J., Eastwood, W., & Perkins, W. (2004). Identification of Aniakchak (Alaska) tephra in Greenland ice core challenges the 1645 BC date for Minoan eruption of Santorini. Geochemistry (as recorded bibliographically), 5. https://doi.org/10.1029/2003GC000672 — opened bibliographically only in this run, abstract not released by the publisher; cited as a title-level claim exclusively.
10. LTRR, University of Arizona, and University of Arizona News (2018). Releases on the work of Pearson et al. 2018. https://ltrr.arizona.edu/news/2018/08/annual-radiocarbon-calibration-re-dates-thera-eruption and https://news.arizona.edu/news/dating-ancient-minoan-eruption-thera-using-tree-rings — opened in this run; source of the details on sample material and sample counts, explicitly not primary literature, and mutually contradictory on one point.

Cross-References in the Vault

• The Solar Storm That Dated the Vikings: 1021 and the Proof Before Columbus — the same toolkit with a luckier anchor: there a single cosmic event supplies a year-precise mark in the wood, here the mark remains disputed.
• Which One of You Was It? The Duhem-Quine Problem and the Underdetermination of Theory — the philosophical foundation: a dating never tests a single hypothesis but always the whole bundle of measurement, calibration and attribution.
• The Year Without a Sun: 536, the Volcanic Winter, and How Ice Cores Redated History — ice cores as a chronology tool, there in the success case; here the case where the attribution of a signal was disputed.
• The Cosmic Tension: Why the Universe Seems to Have Two Rates of Expansion — the same structure in cosmology: two methodologically different measurements of one number that refuse to agree.
• Clocks That Know Their Own Uncertainty: Google Spanner, TrueTime, and Mastering Time in the Cloud — how a distributed system copes with having no shared clock, and why a timestamp with an uncertainty interval is worth more than one without.
• The Networked Collapse: How a Globalized World Fell Apart Around 1200 BC — the same region a few centuries later, and again the question of how to synchronize events across cultural boundaries.
• The Voice of the Clay Tablets: How Linear B Was Deciphered and the Oldest Greek Learned to Speak — the Aegean Bronze Age from the writing side, with the same problem of linking find and time.

← All articles