1. Introduction: sea level as planetary memory

Sea level is a state variable. It integrates the mass of water held in ice sheets, the thermal state of the ocean, and the shape of a solid Earth still relaxing from loads it carried thousands of years ago. Because it integrates, it records: a shoreline is the trace of an ice budget. The terminal Pleistocene rise (conventionally placed near 120 m between the Last Glacial Maximum and the mid-Holocene, though the figure is a reconstruction rather than a reading and different reconstructions state it differently) is the largest coherent environmental signal for the interval in which modern populations occupied every habitable continent. It is planetary memory in the physical sense: information preserved without any agent who intended to preserve it.

There is a second archive, and it is why the subject attracts attention beyond Quaternary science. Communities living on those coasts left narratives, and some concern land that became sea. The temptation to treat the two archives as versions of one another is old and durable. The purpose here is neither to indulge it nor to dismiss it, but to establish what each can bear. The geophysical record is precise about magnitude, increasingly so about timing, and almost silent about experience. The narrative record is eloquent about experience and poor at chronology. That asymmetry is not resolved by placing the two side by side, though much is clarified by saying where each fails.

The article is organised as a two-part reading. The first sets out what is well constrained: the magnitude of the rise, the reason there is no single global curve, and the regionally dated drownings of the Sunda Shelf, the Gulf basin, the southern North Sea and the Black Sea. The second isolates one disputed episode, Meltwater Pulse 1B, because the shape of that dispute is more instructive than its outcome. Placing the contested case in its own section, rather than folding it in, is deliberate. Composite narratives absorb weak claims; separated ones expose them.

Two conventions govern what follows. Every number is given with the uncertainty its source states, and where none is stated the absence is reported rather than smoothed over. Radiocarbon and calendar conventions are distinguished throughout, since the gap between years before 1950 and years before 2000 is fifty years, while that between uncalibrated and calibrated radiocarbon can approach a millennium; a measurable share of the apparent disagreement in the literature on early Holocene flooding is convention mismatch rather than substance. And the standard applied is coherence rather than final demonstration: a reading is preferred when it accommodates more of the evidence with fewer unsupported auxiliary assumptions, and is held provisionally even then.

2. The terminal Pleistocene record: what is well constrained

The modern record begins with a drilled reef. Fairbanks (1989) established the shape of the deglacial rise from uranium-series dated corals off Barbados, converting coral elevations into sea-level elevations by a long-term tectonic uplift correction. That correction is usually quoted as 0.34 mm/yr, or 34 cm/kyr; the Tahiti group's later supplementary material describes the Barbados reconstruction as applying 0.35 mm/yr, characterised as an average of uplift rates measured on different coastal terraces. The averaging is not incidental: an average uplift rate is only as good as the assumption that the cores corrected share one uplift history, the hinge of a dispute discussed below.

Total magnitude is where care is most often abandoned: a figure near 120 m circulates as though it were an observation. It is a derived quantity, dependent on which cores are included, which uplift correction is applied and which model of solid-Earth response is assumed. The extended Barbados record of Peltier and Fairbanks (2006) and the global inversion of Lambeck and colleagues (2014) are the places to source a number, and both present it as model output with stated uncertainty rather than as measurement. Independent shelf stratigraphy agrees: Hanebuth and colleagues (2011) report a Sunda lowstand near −123 m at 20 ka, a figure whose error turns on compaction and subsidence corrections the authors discuss and do not eliminate.

The most important technical point in this literature is that there is no such thing as the sea-level curve. Glacio-isostatic adjustment (the viscoelastic response of mantle and lithosphere to redistributed ice and water, with the gravitational and rotational consequences of that redistribution) makes relative sea level at any site a function of its position with respect to former ice loads and to the migrating water load. The ICE-6G_C (VM5a) model of Peltier, Argus and Drummond (2015) is constrained by space geodesy as well as by geological indicators, a genuine gain in independence, and it remains a model carrying a model's assumptions.

Two consequences follow. Comparing a coral elevation at Barbados with one at Tahiti is not a comparison of two measurements of one quantity; it is a comparison of two site histories, each needing its own correction, so a discrepancy may be a real difference rather than an error in either. And the phrase "a global flood horizon" has no referent in this framework. Different coastlines drowned at different times, rates and depths, and the same ice loss produced different sea-level fingerprints depending on where the ice had been. A narrative of one worldwide event is less contradicted by the geophysics than rendered unformulable within it.

Against that background, one deglacial pulse is robust. Meltwater Pulse 1A appears in the Sunda Shelf record as a rise of "as much as 16 meters within 300 years (14.6 to 14.3 thousand years ago)" in Hanebuth, Stattegger and Grootes (2000), a rate from dated shelf sediments rather than corals, and reported without a formal error term. It appears in the Tahiti boreholes as a resolved step in Deschamps and colleagues (2012), whose title frames it as the Bølling warming rather than a meltwater pulse. And it appears in the Sunda synthesis as a step-like rise of roughly 15 m at shelf depths between −90 and −110 m. Independent archives, independent methods, consistent magnitudes: that is the standard for the next section's episode.

3. Regional drowning, regionally dated

The Sunda Shelf is the largest drowned lowland and the clearest case of the unevenness just described. Hanebuth and colleagues (2011) describe a shelf up to 800 km wide, a lowstand near −123 m at 20 ka and a mean rise of about 10 m/kyr between 19 and 8 ka; major parts of it, they write, "were flooded at the beginning of the Holocene to the average depth of around −50 m". No formal error term accompanies these figures in the published abstract, and the mean rate averages across an interval containing steps, which is a different object from an instantaneous rate.

The palaeogeography accompanying those numbers is more provisional than its reproduction suggests. Voris (2000) produced the standard palaeoshoreline maps at eight contours between 10 and 120 m; they use modern bathymetric contours as proxies for palaeoshorelines, uncorrected for isostasy, tectonics, compaction or later deposition. That is a first-order approximation, not reconstructed palaeotopography, and the distinction matters whenever an exposed-area figure is quoted from such a map. No exposed-area figure for Sundaland is traceable to the primary abstracts surveyed here, and none is asserted. On the mid-Holocene highstand, too, the sources disagree: about +2 m in the 2011 synthesis, +5 m at 4.2 ka in Sathiamurthy and Voris (2006). Both are credible, and the spread is the honest figure.

The Arabian and Persian Gulf basin, mean depth near 35 m, shows why regional curves are necessary. Lambeck (1996) states that sea-level change in the Gulf "is predicted to exhibit considerable spatial variability, because of the response of the Earth to glacial unloading of the distant ice sheets and to the meltwater loading of the Gulf itself". A modest eustatic rise there displaces the shoreline far. The reported chronology has Hormuz opening as a narrow waterway around 14,000 yr BP, incursion into the Central Basin from about 12,500 and the Western Basin about 11,500, and the present shoreline reached about 6,000 years ago, then exceeded by 1–2 m, drowning lower Mesopotamia. These are model predictions dependent on assumed ice history and rheology, as Lambeck says; the radiocarbon-versus-calendar convention is unclear in secondary renderings, where the difference is of order a thousand years; and no shoreline-migration rate is available from this source.

On the same geography Rose (2010) builds a Gulf Oasis hypothesis: an exposed basin watered by the Tigris–Euphrates and by submarine springs, habitable while the peninsula was hyper-arid, whose inundation displaced populations onto the modern coast. The evidential hook is roughly sixty sites appearing on Gulf shores at about 7,500 BP with permanent architecture, pottery, exchange goods and boat remains, but no local antecedents. The load-bearing evidence, though, is submerged: no in-situ dated stratified pre-Ubaid site from the basin floor has been published, which makes the hypothesis hard to test rather than false. Peer commentary published with the article includes a critical response from Hamed Vahdati Nasab, objecting that much of the comparative base is surface collection rather than stratified context, and a supportive one from Juris Zarins. There is also a chronological tension with Lambeck, whose sequence leaves the basin marine well before the site horizon at issue.

Doggerland inverts the usual relation between evidence and narrative. The southern North Sea was mapped as a structured terrestrial landscape from commercial seismic data acquired for hydrocarbon exploration (Gaffney, Fitch and Smith 2009); Sturt, Garrow and Bradley (2013) then modelled palaeogeography at 500-year intervals from 11,000 cal BP, showing spatially uneven inundation and publishing no numeric rates, which should therefore not be supplied on secondary authority. The popular account of how that landscape ended was the Storegga slide tsunami, described by Weninger and colleagues (2008) as catastrophically flooding "large parts of the now submerged North Sea continental shelf" around 8200 cal BP.

That account was withdrawn by the team that had done the mapping. Walker and colleagues (2020), with Gaffney among the authors, report modelled run-up at or below 4 m on the UK mainland and near 7.5 m in maximal Dutch scenarios, against a global mean rise of 1–4 m between 8400 and 8200 cal BP. Their conclusion is that the tsunami "was neither universally catastrophic, nor was it a final flooding event", and that areas of the Dogger Littoral survived "well beyond the Storegga tsunami and into the Neolithic". The event dates given in the two papers differ slightly and should not be run together as one figure.

Ryan and colleagues (1997) proposed that when the Mediterranean rose to the Bosporus sill at 7,150 ± 100 yr BP (a radiocarbon age calibrating to roughly 8,000 cal BP) saltwater "poured through this spillway to refill the lake and submerge, catastrophically, more than 100,000 km²" of shelf. Görür and colleagues (2001) then asked in a title whether the abrupt drowning was a myth; two of its authors had co-authored the 1997 proposal. Aksu and colleagues (2002) read 10–11 kyr of low surface salinity in the Marmara and northern Aegean as persistent outflow. Giosan, Filip and Constantinescu (2009) inferred from Danube delta morphology that "the lake level at the time was around 30 mbsl", not the 80 m or more catastrophism requires. Soulet and colleagues (2011), correcting for a neglected reservoir effect, place the last reconnection near 9,000 cal BP, "in two steps", with a lag of roughly 900 years before lacustrine species vanished.

4. The Meltwater Pulse 1B problem

Meltwater Pulse 1B is the name given to a proposed acceleration of sea-level rise near the end of the Younger Dryas, around 11.3 ka: a second episode of rapid ice loss after the better-attested pulse at the Bølling warming. It has the status of a fixture in review figures and textbook curves. It does not have the status of a robustly observed event. The dispute repays setting out, because it shows how a bounded negative result, a reasserted positive result and a set of specific objections can coexist for years unresolved while the composite literature converges on treating the episode as settled.

Bard, Hamelin and Delanghe-Sabatier (2010) drilled boreholes at Tahiti to test the episode. Their reported finding is that "no significant change in the rate of sea-level rise could be detected" across the interval, "making such an episode unlikely". Their rates, from the supplementary material, are 7.5 ± 1.1 mm/yr during the Younger Dryas against 12.1 ± 0.6 mm/yr before and 11.7 ± 0.4 mm/yr after; the Barbados figure for the stadial is 5.6 ± 0.4 mm/yr. Those errors are the authors' own, reflecting age-model and depth uncertainties within a coral archive; they do not propagate the glacio-isostatic correction needed to convert coral elevations into sea-level elevations, a separate and larger term.

The qualifier the Tahiti authors supply is the most important sentence in the exchange, and it is routinely dropped. A step of the proposed kind, they note, would be "masked by the inherent uncertainty linked to the coral approach (less than 6 m)". Tahiti therefore excludes a step of about 15 m and does not exclude a step below roughly 6 m. This is a bounded negative result, not an absence of the thing. To say Tahiti rules the episode out misstates the finding in the direction of strength; to say it is silent misstates it in the direction of weakness. The accurate formulation is narrow: no rate change is detected, and any step is constrained below the method's resolution.

One feature of that negative result forecloses the obvious rejoinder. The same Tahiti programme, on the same archive and broadly the same methods, resolved Meltwater Pulse 1A in well-defined form, and the Tahiti corals had already supplied the deglacial record of Bard and colleagues (1996). An archive registering one pulse and not another cannot be dismissed as insensitive to pulses in general. This does not settle the question, since a record could still miss a pulse of particular size or origin. It does remove the cheapest objection, and it is why the 2010 result carries the weight it does.

The positive claim was reasserted six years later by the original group. Abdul, Mortlock, Wright and Fairbanks (2016) report from Barbados reef-crest Acropora palmata that rates declined smoothly from 20 mm/yr to 4 mm/yr between about 13,900 and 11,650 yr BP, and that after a brief stasis, from 11,450 to 11,100 yr BP, "sea level rose 14 ± 2 m and rates of sea level rise reached 40 mm yr⁻¹", read as a delayed ice-sheet response to North Atlantic warming with a lag near 400 years. The stated ± 2 m is an uncertainty on the amplitude of a step whose existence is the point at issue; it does not propagate the disputed uplift correction, and is not a total error.

Bard, Hamelin, Deschamps and Camoin (2016) responded with three objections whose specificity makes them assessable. The first is tectonic. Barbados sits above a subduction zone where uplift is neither uniform nor well known, and the cores defining the step "are drilled in different environments: submerged fossil barrier reef for the deepest cores #12 and 16, and fringing reef for the shallow core #7", possibly on different tectonic segments with different uplift rates. On that reading the step is a splice artefact between cores needing different corrections, the terrace-averaged uplift rate noted earlier. The second objection concerns the indicator species. Acropora palmata is "generally considered as representative of water depths less than 5 m", and the Barbados interpretation requires that "during the sea-level transition, the depth tolerance of A. palmata exceeded its 5-m limit". A reef-crest species that lagged a rapid rise, then caught up, would manufacture a step from a smooth rise.

The third objection is glacio-isostatic adjustment. Because Barbados and Tahiti occupy different positions relative to former ice loads and to the migrating water load, their sea-level histories are expected to differ, so a step present at one and absent at the other is not automatically a contradiction. Bard and colleagues' supplementary material concedes the reconciliation this permits: an apparent 15 m step at Barbados "may be reconciled by assuming that the ice was released exclusively from the Pacific sector of the Antarctic ice sheet", a source whose fingerprint is large at Barbados and small at Tahiti. That is the honest middle of the dispute, and an auxiliary assumption with its own burden, needing Antarctic mass-balance support the sea-level data cannot supply.

The defensible summary is layered rather than binary. Meltwater Pulse 1A is reproduced across independent archives and methods with consistent magnitudes. Meltwater Pulse 1B is not: present in the Barbados corals, undetected in the Tahiti boreholes, and challenged on differential uplift between spliced cores, on the indicator species' depth behaviour during rapid rise, and on the impossibility of direct inter-site comparison. Sedimentological and geophysical work continues to address it, which weighs toward the question being live rather than closed. What is not supportable is treating the pulse as an established event.

5. Chronological placement: the Younger Dryas and its misquoted speed

The disputed pulse sits at the Younger Dryas termination, so that stadial's chronology matters. The controlling framework is the INTIMATE event stratigraphy of Rasmussen and colleagues (2014), on three synchronised Greenland ice cores: Greenland Stadial 1 begins at 12,896 b2k with a maximum counting error of 138 years and ends at 11,703 b2k with one of 99 years, a duration of 1,193 years. Two conventions must be held apart: b2k counts from 2000 CE and cal BP from 1950 CE, placing those boundaries at 12,846 and 11,653 cal BP. The ± 4 years often quoted with such dates is the precision of transition placement, not the age uncertainty. Walker and colleagues (2009) define the Holocene base at 11,700 b2k in the NGRIP core, with the same 99-year error, on "an abrupt shift in deuterium excess values".

Deuterium excess is a second-order isotopic parameter, not a Greenland temperature proxy. In Steffensen and colleagues (2008) it is "mainly a proxy of past ocean surface temperatures at the moisture-source region", where the snow came from, not how cold Greenland was. The celebrated figure of one to three years for the end of the Younger Dryas is a deuterium-excess transition, corresponding to a shift of 2–4 K in moisture-source temperature, and so to an atmospheric reorganisation. From the durations tabulated there, the δ¹⁸O warming at the termination took 60 years, the deuterium-excess shift 3. The paper states that the warming at 11.7 ka "lasted 60 years", both transitions corresponding to "a warming of more than 10 K". Sixty years remains extreme, and it is the figure referring to temperature.

The onset is more discrepant still. The same tabulation gives a δ¹⁸O onset transition of about 213 years, with ramp-fitting errors of ± 59 and ± 74 years, against a deuterium-excess transition of about one year: onset and termination are strongly asymmetric, slow in, fast out. The asymmetry is reproduced in an independent archive. Cheng and colleagues (2020), from absolutely dated speleothems, place the onset at 12,870 ± 30 BP by uranium–thorium and describe "an abrupt drop in δ18O (∼2‰ within ∼20 y)" inside a roughly 200-year excursion, a sharp inflection within a slow transition. A reflexive observation comes free: the boundaries are pinned to the fastest-switching proxy, which is why the canonical dates look sharper than the history they date.

Magnitude requires a proxy independent of δ¹⁸O, and firn thermometry on δ¹⁵N and δ⁴⁰Ar supplies one. Buizert and colleagues (2014) report onset cooling of −4.8 ± 1.2 °C at NEEM and −9.2 ± 0.9 °C at GISP2 in central Greenland, with terminal warming of +8.4 ± 3.0 °C and +12.4 ± 3.4 °C; the uncertainties are two standard deviations, and the authors flag "an unexplained abrupt shift in the d15N data" at NGRIP, which is why that site is not used here as a bracket. The same analysis reports the Younger Dryas as 4.5° ± 2 °C warmer than the Oldest Dryas, attributed to carbon dioxide forcing and summer insolation, and anticipated in coupled-model work. Any magnitude read from a δ¹⁸O curve alone is biased by that offset.

Cross-archive agreement is now good. Converted to a common scale, onset estimates from the ice-core count, the speleothem chronology and the redated Laacher See tephra span roughly 63 years, less than the 138-year counting error on the ice-core boundary. Reinig and colleagues (2021) redated that eruption to 13,006 ± 9 cal BP and reported that the result "synchronize[s] the onset of the Younger Dryas across the North Atlantic–European sector", removing much of the regional diachroneity earlier literature inferred; Warken and colleagues (2025) later found it in a speleothem record. A small discrepancy remains and should be stated rather than averaged away: Reinig's figures imply the eruption preceded the onset by 199 years, Warken's by about 150. Both treat the residual as within uncertainty, and the two numbers are not interchangeable.

6. Reading memory traditions as data fragments: method and its limits

The most developed attempt to read narrative traditions against a sea-level curve is that of Nunn and Reid (2016), who argue that Aboriginal Australian accounts of coastal inundation preserve memory of postglacial marine transgression, and are therefore of an order of age unusual in the study of oral tradition. The argument has a clean form: identify narratives describing named locations now submerged, establish their depth, and read the corresponding date from a regional sea-level curve. Where the geography described is unambiguous and the bathymetry known, the inference is at least well posed. It is also, as the authors are more candid about than their reception suggests, wide.

Two features of the paper should be reported carefully. Its own tally of narratives is quoted inconsistently in the citing literature: eighteen coastal stories in the authors' own popular account of the work, twenty-one locations in many secondary references. A review of this kind is not positioned to adjudicate between those figures and prints neither as authoritative, a small case of how a number acquires currency. The second is the age range the authors give. The narratives could be from 6,000 to 13,000 years old depending on which sea-level datum they encode, with more conservative interpretations placing some near 10,000 years, and the authors acknowledge uncertainty over whether particular details recall glacial conditions rather than later periods.

That range is not vagueness. It is the direct expression of an inference structure that is circular in form. The narratives are dated by matching their content to a sea-level curve; they are then offered as evidence that memory persists across the interval so derived. Nothing independent fixes the datum, and the range is wide precisely because the depth encoded is inferred from the story rather than measured against it. An independent anchor (an archaeological horizon, a dated volcanic event, a linguistic stratum with its own chronology) would break the circle, which is why the extension by Hamacher and colleagues (2023) to Tasmanian traditions, and Nunn's earlier synthesis (2016), matter more for method than conclusions.

On mechanism the authors propose that transmission of this fidelity is supported by a conservative and explicit approach to "the law" governing knowledge, by high cultural value placed on preserving information accurately, and by kin-based systems permitting cross-generational cross-checking, together an "inter-generational scaffolding". This is a proposal about social structure rather than an appeal to unexplained persistence, testable in principle against ethnographic description. It should be recorded that no formal published rebuttal, comment or reply to Nunn and Reid appears in the indexed journal record; the literature that has followed extends the claim rather than contesting it. That is a different situation from a contested claim, indicating an absence of adversarial testing rather than the survival of a claim that has been tested.

Indexed databases cover journals well and books unevenly, so the accurate statement is that no formal reply appears in the journal record, not that no criticism exists anywhere. Three structural cautions stand on their own. The recording context is late, these narratives having been written down in the nineteenth and twentieth centuries largely by non-Indigenous collectors, raising the feedback problem long recognised in oral historiography. Coastal change is a recurring rather than unique referent, since storm surge, subsidence, dune migration and river-mouth change all generate narratable inundation. And the classical literature sets shorter reliable horizons: Jan Vansina's floating gap describes detailed recent events and origin narrative separated by a hollow middle, while David Henige is sceptical about deriving chronology from oral sources. Neither yields a horizon quotable as a number.

7. What apparent similarity might actually be

A recurring move in this area is to establish the significance of flood narratives by counting them, and the totals that circulate are seldom traceable to a defined sample. The difficulty is not that a count is impossible but that a count is definitionally dependent in a way no amount of decimal precision repairs. Whether a narrative counts as a flood tradition depends on whether one requires a named survivor, a divine cause, a vessel, a global rather than local inundation, a post-diluvial repopulation, or simply water covering land. Each criterion yields a different sample from the same corpus, and the figures are not comparable. A bare number therefore cannot be authoritative, and the appearance of authority it confers is the hazard.

The natural place to seek a defensible count is the largest systematic compilation in the field, the analytical catalogue of world mythology developed by Yuri Berezkin for studies of areal motif distribution. It is a motif index, organised by motif and area for distributional analysis. It is not a hypothesis-testing dataset with a coding scheme, an inter-rater reliability procedure and a defined population, and does not present itself as one; any count extracted from it inherits whatever definition the extractor supplied. The canonical interpretive compilations are likewise essay collections, not datasets: the volume edited by Alan Dundes (1988) shaped several fields and was reviewed across folklore, anthropology, classics and religious studies, interpretive influence, not coding. Quantitative comparative mythology does exist (phylogenetic methods have been applied to folktale distributions, by da Silva and Tehrani (2016) on Indo-European tales) but none of those published analyses treats flood narratives, and borrowing a method's credibility without its results produces fabricated statistics.

There is a model of what a properly bounded comparative claim looks like, and it concerns a flood motif directly. Korotayev, Berezkin, Kozmin and Arkhipova (2006) examined the post-diluvial bird-reconnaissance motif (the bird sent out after the waters to find land) and its White Raven variant, indexed as Thompson motif A2234.1.1, testing its areal distribution against Anna Birgitta Rooth's proposal that the motif reached the New World with Catholic missionaries. Their conclusion: "both 'The Raven and the Carcass' motif and its 'White Raven' version existed in the New World prior to the European colonization of Americas". That is a claim about one motif, one transmission route and one terminus, silent about antiquity of any greater depth. The restraint is what makes it usable.

The strongest self-critical move here is not about counting but about where similarity comes from. Michael Witzel's The Origins of the World's Mythologies (2012) proposes a deep division between a Laurasian mythological complex, marked by a unified narrative arc from creation to destruction, and a Gondwana complex lacking it. Reviewing that book, Bruce Lincoln (2015) states his belief that Witzel "has written a seriously flawed book whose conclusions carry racist implications", objecting to reliance on "dated German literature steeped in the Kulturkreis paradigm" and on "works written by poets, rather than ranking authorities". He also quotes Witzel's own concession that "Even a certain accumulation of circumstantial counterevidence does not suffice to bring down the theory", a statement about the theory's relation to evidence rather than about the evidence.

Lincoln's central formulation is the one this article puts to work: "The contrast is not between 'Laurasian' and 'Gondwana' mythologies, but between indigenous and exogenous agents, processes, and products of textualization." Where an insider scholar produced a unified written narrative (Snorri Sturluson, Hesiod) a reader finds a story line; where an outside collector assembled an anthology, a reader finds fragments. The difference lies in the style of collection rather than in the mythologies, and it manufactures the pattern the thesis requires. The implication generalises beyond Witzel: apparent structural similarity between widely separated flood traditions may be, in part, a property of how they were collected, translated and edited into comparable objects. Where Mesopotamian traditions and their Near Eastern recensions survive as continuous written narratives while others survive as collectors' fragments, resemblance among the written ones is not straightforwardly evidence about the traditions.

8. Limits, alternative readings, and what would change the reading

Three readings of Meltwater Pulse 1B remain open, less equally probable than differently costly. On the first, the step is an artefact of splicing cores from different reef environments and possibly different tectonic segments, compounded by an indicator species lagging a rapid rise; this needs no new physics and is cheapest, but must explain why the effect yields a coherent step rather than noise. On the second, it is real and regional, from the Pacific sector of the Antarctic ice sheet; this accommodates both records at the cost of a mass-balance commitment the sea-level data cannot discharge. On the third, it is real and global but smaller than the roughly 6 m Tahiti resolves, in which case Tahiti's silence is expected and the Barbados amplitude of 14 ± 2 m is overstated.

Each reading implies a different decisive observation, which makes the dispute tractable rather than merely persistent. Redating the Barbados cores against segment-specific uplift rates, rather than an average across terraces, would test the first, and the tectonic constraint required is geodetic and geomorphological rather than palaeoclimatic. A far-field record with sub-metre resolution across 11.45–11.1 ka, from an archive not dependent on a reef-crest species' depth tolerance, would test the third. Antarctic mass-balance evidence bearing on the Pacific sector across that interval would test the second, and it is the only one of the three whose test lies outside sea-level science.

On the narrative side the analogous requirement is an independent date, an anchor that does not come from the story. A published critical response to Nunn and Reid, of the kind absent from the indexed journal record, would also change the reading, and its absence is a gap in the evidential situation rather than tacit endorsement. An alternative reading deserves its strongest form: traditions of land becoming sea may encode the ordinary, repeated experience of coastal change, later generalised and given cosmological weight, which accounts for their geographical breadth more economically than deep transmission does, and for named submerged locations at appropriate depths less well.

The regional cases carry a parallel lesson. In every instance here the direction of revision has run the same way: the Black Sea from a breach to a two-stage reconnection spanning ten centuries, Doggerland from a terminal wave to a littoral persisting into the Neolithic, the Gulf basin from a single inundation date to a model-dependent sequence. In no instance has the revision amounted to nothing having happened. Large areas of shelf were submerged, populations were displaced, and at human scale the events were large. The edited volume assembled by Yanko-Hombach and colleagues (2007) is useful precisely because it preserves incompatible positions rather than resolving them prematurely.

9. Open questions

Several questions are open in the specific sense that a stated observation would settle them. Whether the Barbados cores defining Meltwater Pulse 1B share one uplift history is a determinate question about tectonic segmentation, currently answered by an average across terraces. Whether a step below roughly 6 m occurred is answerable in principle by an archive with finer resolution than corals at Tahiti provide, though the far-field constraint making such a record decisive is a modelling problem as much as a drilling one. And whether the 2016 exchange closed with the comment or continued into a formal reply is a bibliographic matter this review does not resolve.

On chronology, the residual disagreement over the interval between the Laacher See eruption and the Younger Dryas onset (199 years on one set of figures, about 150 on another, both described as within uncertainty) is small but not empty, and bears on how tightly European terrestrial records can be tied to the Greenland count. A second question concerns the count rather than the climate. If the ice-core chronology near the Younger Dryas is better than its formal maximum counting error, as the speleothem comparison has been read to suggest, the observed spread of roughly 63 years between archives means something different from what it means under the formal error.

Regionally, the most consequential gap remains the absence of an in-situ dated stratified pre-Ubaid site from the floor of the Gulf basin. Its absence is expected on the hypothesis it would support, which is what makes that hypothesis hard to test, and no modelling substitutes for a dated context. Comparable gaps are of the same kind: no exposed-area figure for Sundaland traceable to a primary reconstruction rather than to a bathymetric proxy, no published numeric inundation rates for the North Sea models, and a mid-Holocene Sunda highstand reported at about +2 m and at +5 m by credible sources whose difference has not been reconciled.

On the narrative archive the open questions are methodological before they are empirical. What would constitute an independent date for a narrative, as against a match between its content and a curve, has not been settled in a form the field agrees on. Whether the fidelity mechanisms Nunn and Reid describe can be assessed against ethnographic description independently of the conclusions they support is likewise unresolved. And Lincoln's textualisation argument has not, so far as the present survey establishes, been applied systematically to flood traditions in particular, which is the study this article would most like to read. Until it exists, coherence remains the measure, and coherence is not demonstration.

1. Introduction: a site that unsettled a standing narrative

For most of the twentieth century the standard Neolithic sequence ran in one direction: subsistence change, then sedentism, then surplus, then monuments. Göbekli Tepe, on a limestone ridge in Şanlıurfa province, south-eastern Turkey, disturbed that order. Excavation from 1995 exposed sunken enclosures framed by T-shaped limestone pillars, the largest standing 5.5 m and estimated at eight to ten metric tons apiece (Dietrich et al. 2012), in deposits associated with hunted fauna and wild plants. That construction begins in the tenth millennium cal BC. Whatever else the site was, it was not the product of an agricultural surplus.

The rhetorical consequences arrived quickly. Schmidt's formula, zuerst kam der Tempel, dann die Stadt, first came the temple, then the city: was published as a preliminary report title (Schmidt 2000) and has circulated since as though it were a result. It is better described as a hypothesis about causal direction, and it has drawn sustained objection. This review neither defends nor dismisses the slogan; it sets out what the evidence constrains, and is explicit where the excavation team has withdrawn positions on which earlier syntheses, including an earlier version of the reading offered here, depended.

Two such revisions are central. Clare (2020) records the abandonment of Schmidt's tripartite chronological sequence, and in the same report replaces the intentional backfilling of the enclosures with a taphonomic model of slope-derived inundation accompanied by prehistoric mitigation. Both had circulated very widely; both are now behind the literature. A third shift is not a retraction but an accumulation: the sanctuary-versus-settlement opposition that structured debate for two decades has largely dissolved, less through argument than through the recovery of domestic evidence at the site itself. Any synthesis written before about 2020 needs checking against these three developments.

This article proceeds on a stated principle: coherence is the measure, not final demonstration. Nothing below is offered as proof. Dates are given with the method that produced them and with the caveats their authors attached; where a claim rests on a field report rather than a peer-reviewed research article, or on press coverage rather than publication, that is said. Calibrated determinations are reported as cal BC throughout, except where a cited author reports uncalibrated years before present, in which case the original convention is retained and identified as such.

The reading proposed at the end, the site as residual expression, is a reading. It is advanced because it accommodates the revised chronology better than the sequential narrative it replaces, not because the evidence selects it over rivals. The closing section sets out what would count against it. That is the appropriate ambition for a site where much of the mound remains unexcavated and where no comprehensive radiocarbon monograph or published Bayesian chronological model yet exists. What follows is a review of constraints, and of the places where those constraints have lately moved.

2. Chronology and how it is known

The span currently given for the site is approximately 9500–8000 cal BC (Clare 2020). Within it, the Pre-Pottery Neolithic A falls at roughly 9500–8700 cal BC and the Early Pre-Pottery Neolithic B at roughly 8700–8200 cal BC, conventional regional brackets, not site-specific boundaries. The radiocarbon series behind them is modest, drawn from heterogeneous materials, and has not been synthesised in a chronology monograph or a published Bayesian model. The revised chronology rests principally on an eight-page field report (Clare 2020) and a later synthesis article (Clare 2024): a real limit on how firmly any of it can be phrased.

The determinations themselves show why caution is warranted. Wall plaster from Enclosure D (KIA-44149) returned 9984 ± 42 ¹⁴C BP, calibrating to 9745–9314 cal BC at 95.4 per cent (Dietrich and Schmidt 2010; Dietrich et al. 2013). Charcoal from the Enclosure D deep sounding (UGAMS-10795/96/99) gave 9664–9311 cal BC at the same confidence; charcoal from the Enclosure C deep sounding (UGAMS-10797) returned 9700 ± 30 ¹⁴C BP, calibrating to 9261–9139 cal BC at 91.6 per cent. Animal tooth collagen from Enclosure D fill (KIA-44701) gave 9800 ± 120 ¹⁴C BP but calibrated across 9746–8818 cal BC: a window of nearly a millennium.

Each material carries its own liability, stated plainly by the excavators. Organic remains from fill sediments "could be older or younger than the enclosures" (Dietrich et al. 2013). Pedogenic carbonate coatings accumulate at a variable rate, yielding average values rather than event dates. Bone suffers from poor collagen conservation and from carbonate-rich sediments that complicate apatite fractions, and a calibration-curve plateau widens several probability ranges regardless of sample quality. On relative order the same restraint applies: the "still limited series of radiocarbon data seems to suggest" that the Layer III enclosures were not exactly contemporaneous, Enclosure D earliest and Enclosure A younger than C and D. The result is a set of loose constraints, not a timeline.

The methodological development reported by Clare (2020) is the direct dating, by accelerator mass spectrometry, of organic residues extracted from the mud-mortar of the buildings themselves. That targets the construction event rather than its surroundings, and it is what undid the layer scheme: "multi-phase structures were much longer-lived than previously acknowledged; indeed, in some cases their final building phases are consistent with a construction date around the mid-ninth millennium calBC." Round-oval special buildings and rectangular structures therefore overlap in time rather than succeeding one another, and enclosures built at different times cannot be read as a single standing complex.

Clare states the consequence without softening it: "This realization has led to the abandonment of Schmidt's tripartite chronological sequence for Göbekli Tepe which also included an uppermost Layer I (plough-horizon)." The labels Layer II and Layer III remain serviceable as shorthand for building types (round-oval with large pillars, rectangular with smaller ones) and are used in that restricted sense below. They no longer name stratigraphic phases. Several works cited in this review predate the revision and still treat them as a succession; readers should adjust accordingly rather than assume the older framework has been quietly retained.

3. Anatomy of the T-pillars

The pillars are monolithic limestone, T-shaped in elevation, set in sockets cut into bedrock or prepared floors. Peters and Schmidt (2004) give average heights of 3.5–5 m. The central pair of Enclosure D stands 5.5 m at an estimated eight to ten metric tons each; surrounding pillars in the same enclosure approach 4 m, and Enclosure C's original central pillars around 5 m (Dietrich et al. 2012). An unfinished example on the western plateau measures roughly 7 m by 1.5 m, with an estimated mass near 50 metric tons, an intention rather than an achievement.

The anthropomorphic reading originates with the excavators and is stated carefully. Peters and Schmidt (2004) write that "the T-shaped pillars thus seem to represent stylised anthropomorphic beings, the horizontal and vertical parts respectively being the head and body." The support is specific: some monoliths carry a pair of arms and hands in bas-relief, and parallel grooves have occasionally been recorded which "probably" refer to human clothing. The hedge is the authors' own. Confident descriptions of belts and loincloths run firmer than the primary literature warrants, and the reading remains an inference from schematic form rather than a recovered identity.

The relief repertoire is better quantified: at least ten vertebrate taxa on 22 of 37 pillars examined by Peters and Schmidt (2004). Snakes account for 28.4 per cent of representations, foxes 14.8, wild boar 8.7, cranes 6.2 and aurochs 3.7 per cent. That distribution does not track diet. Of 38,704 mammalian remains, ungulates supply more than 80 per cent of identifiable specimens, and goitred gazelle is the most hunted animal at 41.8 per cent of NISP (n = 7,949); yet aurochs "provided about 50% of the total meat consumed, whereas gazelle … only contributed some 15%."

The gap between what was eaten and what was carved is among the site's more durable observations, and the excavators decline to close it. They read the bone assemblage as "refuse derived from hunting and food preparation … rather than from ritual procedures," note that "it is difficult to believe that architecture and iconography at Göbekli only served for hunting rituals," and conclude that the purpose of the enclosures "will take much more time to be properly answered." Dietrich and Notroff (2016) extend the caution to method, publishing a decorated bone object as a case study in the pitfalls of reading early Neolithic imagery.

Two further bodies of evidence bear on the pillars without settling them. Gresky, Haelm and Clare (2017) describe modified human crania carrying deliberate incisions and, in one instance, drilling, which they read as a previously undocumented form of Neolithic skull treatment; the headless figure on Pillar 43 has been interpreted against that background and against documented mortuary practice involving severed heads and necrophagous birds (Notroff et al. 2017). Dietrich, Notroff and Dietrich (2018) propose masking practices from stone objects at the site and its neighbours. Both remain interpretive proposals on modest evidential bases.

4. Enclosures as planned space, and the unexcavated remainder

The enclosures are sunken, roughly circular to oval spaces, in the larger cases some ten to twenty metres across, with pillars set into a perimeter wall and a free-standing central pair. Kurapkat (2012) argues from structural evidence that at least some were roofed, a position Banning (2023) records as having "gained traction." Roofing is not a minor detail. It constrains what the interiors could have been used for, it bears on heating and access, and it bears directly on astronomical readings, since a covered enclosure cannot function as an observatory.

What lies beyond the excavated area is known chiefly through geophysics. Magnetic prospection was applied extensively, with georadar and geoelectric tomography in selected zones, across campaigns in 2003, 2006, 2007 and 2012. A 2003 field report by Schmidt recorded more than ten large enclosures in the geomagnetic map, with more to be expected; adding the four then under excavation gave an estimate of some twenty enclosures and, at roughly twelve pillars each, more than two hundred pillars. Dietrich et al. (2012) restate the judgement as "it seems very probable that at least 20 enclosures existed in total."

Two qualifications attach to that figure, and any account repeating it without them misleads. The first is the project's own: only very large structures register clearly in the survey maps, so the estimate is a lower bound for big buildings and silent about everything smaller. The second follows from the revised chronology. If the enclosures were neither contemporaneous nor sealed episodes, twenty enclosures does not describe twenty structures standing together; it describes a cumulative inventory across some fifteen centuries. The image of a precinct of twenty temples has no support in the geophysics.

Against this, Haklay and Gopher (2020) argue for deliberate planning above the scale of the single building: on geometric grounds they propose that three of the large stone-built enclosures "were planned and initially built as a single project." The claim is structural rather than chronological, it is contested, and it depends on which pillar positions are original: many are not, since pillars were reused and repositioned during the life of the site (Notroff et al. 2017). Whether the enclosures represent a designed ensemble, a sequence of emulations, or both at different moments remains open, and the unexcavated remainder is where any answer sits.

5. The burial question, revised

For two decades the standard account of the mound's formation was intentional. Dietrich et al. (2012) state it flatly: "The mound is the result of the rapid and intentional backfilling of these circles after some time of use." The claim was attractive on several grounds. It explained the survival of the pillars in near-original position, it connected Göbekli Tepe to a wider Pre-Pottery Neolithic practice of building closure, and it supplied a ritual reading of the site's end as well as its beginning. Schmidt advanced it repeatedly from 2000 onward, and it passed into general circulation as settled fact.

Clare (2020) withdraws it. Earlier work, he writes, "postulated an intentional (ritually-motivated) backfilling of these structures at the close of their active use-lives"; but "a review of the available evidence now suggests an inundation by displaced archaeological deposits from adjacent higher-lying knolls and slopes of the mound." The replacement is not a bare appeal to nature. It "explains the source and the catchment of the debris, which as we now know contains mixed PPNA and PPNB materials, and the heterogeneity of previous radiocarbon measurements", two independent anomalies that the ritual model had left unaccounted for.

What changed the reading, then, was internal consistency rather than a single new discovery. Mixed Pre-Pottery Neolithic A and B material within the fills is hard to reconcile with rapid deliberate burial at the close of use, and inconsistent radiocarbon results from those same fills point the same way. Slope failure delivering deposits from higher ground accommodates both. Clare adds that "details of the multiple slope slide events, as well as mitigation efforts undertaken by the prehistoric population to protect the special buildings, are an area of continued research." Human agency stays in the model, redirected from burial to protection.

Two cautions follow. The first is against overcorrection. Deliberate building-burial is well attested in the Pre-Pottery Neolithic generally, has been debated since the 1960s, and is argued directly for a neighbouring site: Karul (2021) reports that "all buildings excavated at Karahantepe appear to have been deliberately filled." Kuijt (2008) sets such practices within a broader framework of structured remembering and forgetting. What has been withdrawn is the Göbekli Tepe attribution, not the regional practice. The distinction is between a documented regional practice and one particular application of it, an application now retracted by the people who proposed it.

The second caution concerns the state of the abandonment literature. The dedicated studies are those of Schönicke (2018, 2021), whose work is sometimes misattributed to other members of the project, and whose 2021 chapter argues towards a revised understanding of abandonment practices at the site; Kinzel and Clare (2020) and Kinzel (2024) treat the architectural dimension. A field that publishes the reversal of its own influential position in an eight-page report, and leaves the detailed taphonomy to continuing work, is behaving properly but has not closed the question. The current position is best described as a mixed taphonomic model under active revision.

6. Sanctuary or settlement: a dissolved dichotomy

Banning (2011) applied the strongest early pressure to the temple reading. His argument was partly evidential and partly conceptual: "the evidence supporting the identification of cult buildings is often equivocal or depends on ethnocentric distinctions between sacred and profane spaces," and "consideration of the evidence suggests the alternative hypothesis that the buildings at Göbekli Tepe may actually be houses, albeit ones that are rich in symbolic content." The paper appeared with published commentary from Baird, Goring-Morris and Belfer-Cohen, Hauptmann, Hodder, Özdoğan and Rosenberg; a spread of reaction that remains a useful index of where the field then stood.

The excavators replied in detail (Dietrich and Notroff 2015). Their case against habitation was cumulative: the nearest springs lie some 5 km away and cistern capacity appeared insufficient for a rainless summer; no hearths were found inside the enclosures; high interiors would be impractical to heat; the summit is exposed; and object classes typical of domestic assemblages, clay figurines among them, were absent. They contrasted the site with Çayönü and Nevalı Çori, where domestic architecture is unambiguous. It was a serious defence, and several of its individual points have since been overtaken by excavation at Göbekli Tepe itself.

Clare (2020) reports round-oval structures in an agglutinative arrangement with a larger multi-phase round-oval building, bordering "a small activity area that produced a sequence of hearths and evidence for stone and bone bead production with unusually high frequencies of bone tools." The lithic assemblages have compositions "wholly characteristic of domestic contexts." On water: a partially excavated rock-carved pit 8.0 m across and 2.8 m deep, plus carved channels and rock-cut cisterns indicating rainwater harvesting from the adjacent plateau. A kidney-shaped sub-floor cut in a rectangular building held the disturbed remains of at least three individuals, "a common feature of PPN domestic sites."

Food processing supplies the largest single body of evidence. Dietrich et al. (2019) analysed more than 7,000 ground-stone artefacts (grinding slabs and bowls, handstones, pestles, mortars) of which some 3,400 are handstones, about three-quarters belonging to two types optimised for cereal grinding. Phytolith analysis indicates the massive presence of cereals, with grass phytoliths forming roughly 81 per cent of sampled material. The authors find "a clear link between plant food preparation and the rectangular buildings" and delimited working areas on the terraces. Laura Dietrich's 2021 monograph presents the tool corpus in full.

The dichotomy has not been resolved in favour of one side; it has stopped being productive. Banning (2023) notes that "some elements of that critique, including assertions about roofing and a residential population, have gained traction, while others have fallen on deaf ears." Schönicke (2018) describes the site, from within the project, as "nine hectares densely covered with round and rectangular buildings with sets of domestic activities of daily life." Özdoğan (2024) states the regional point most clearly: special buildings at Sayburç stand "close to dwellings" and illustrate "the intertwining of the sacred with everyday life." Jeunesse (2020) and Hodder (2020) offer alternatives that likewise decline it.

7. Taş Tepeler: the site is not isolated, and the record is uneven

Göbekli Tepe belongs to a cluster of sites with T-shaped pillars across the Urfa plain and its margins, now grouped under the Taş Tepeler research programme. The distribution has been known in outline since the late 1990s through survey, and the project's own 2016 review of pillar sites recorded a constraint that still holds: the large pillars, at 4–5.5 m, were then known only from Göbekli Tepe. Elsewhere the pillars are smaller, comparable to those of Göbekli's rectangular buildings. The record is uneven in a way that should govern how confidently each site is invoked. Nevalı Çori, excavated between 1983 and 1991 and now flooded by the Atatürk reservoir, is reported in Hauptmann's work (1988, Anatolica 15: 99–110). Its pillars stand in an unambiguous settlement containing a cult building, precisely why the Göbekli excavators long used it as a contrast case.

Sayburç is the best-published of the recently excavated sites, with two journal outputs (Özdoğan 2022, 2024). The 2022 item is an Antiquity Project Gallery report (editorially reviewed rather than peer-reviewed) describing a relief panel of roughly 0.7–0.9 by 3.7 m on a bench within an 11 m circular building: a central male figure in high relief flanked by two leopards, a second squatting male with a bull whose head is rendered from above, phalluses displayed, one figure wearing a triangular necklace, one holding a snake or rattle. It is presented as the earliest known narrative depiction: a claim about relative dating as much as about what counts as narrative.

Karahantepe is the site most often invoked and least securely published internationally. Survey began in 1997 and excavation in 2019 under Karul, whose 2021 paper on buried buildings is the substantive accessible statement; a water-systems study (Polat et al. 2023, Environmental Archaeology 28 [6]: 475–487) is peer-reviewed but narrow. The much-reproduced human-head pillar and the chamber of phallic pillars are, so far as could be verified, press-reported rather than peer-published. Harbetsuvan Tepesi, excavated in 2017–18, appears in a Turkish-language regional journal and is described by its excavator as a secondary cult centre. Sefer Tepe, Kurt Tepesi and Çakmaktepe have no verified peer-reviewed excavation reports.

The asymmetry is itself informative. The Taş Tepeler programme has transformed the picture of the region faster than it has transformed the literature about it, and much of the current public understanding of Karahantepe rests on material that has not passed peer review. This is not an accusation; excavation outpaces publication as a matter of course, and the programme is young. It does mean that comparative arguments built on the new sites carry the evidential weight of preliminary reporting, and that treating these sites as equally documented levels a real difference in standing.

8. Feasting, labour and social scale

Dietrich et al. (2012) assembled the case for consumption at large scale. Six barrel- and trough-shaped limestone basins from Pre-Pottery Neolithic B contexts, with capacities up to 160 litres, are described as static, integral parts of particular rooms. Residue testing produced "probable evidence of oxalate for some samples," a compound that "develops during the steeping, mashing and fermentation of cereals", and the same paper records that the testing "proved inconsistent." The faunal component is less equivocal: large quantities of animal bone smashed for marrow, in amounts that "exceed everything known from contemporary settlements."

The fermented-beverage reading should be left open rather than asserted or denied. The 2012 evidence is tentative by its authors' own account. An integrated re-examination of the troughs was published in 2020 (Dietrich et al., Journal of Archaeological Science: Reports 34: 102618) by a large specialist team including the leading residue analyst in the field; its content could not be consulted for this review, and it is the natural place for the question to be resolved in either direction. Until it has been read, brewing at Göbekli Tepe is a hypothesis with partial and internally inconsistent support.

The work-feast interpretation stands on firmer ground, resting on a different argument. Dietrich et al. (2019) observe that no large storage facilities have been identified, and therefore "argue for a production of food for immediate use and interpret these seasonal peaks in activity at the site as evidence for the organization of large work feasts." Absence of storage combined with intensity of processing is a reasonably tight constraint, though it rests on a negative and would not survive the discovery of granaries. Hayden (2014) situates the model within debates on competitive feasting before cultivation; Dietrich, Notroff and Schmidt (2017) develop it for Upper Mesopotamia.

On labour the honest statement is that no quantitative estimate exists: no published workforce or person-hour calculation for Göbekli Tepe could be located. The figure that circulates, several hundred men hauling a megalith some kilometres in a few days with a wooden sledge, is an ethnographic analogy from island Southeast Asia, cited by Dietrich et al. (2012) as a comparison, not as a computation for this site. What the project does supply is a scale contrast: core group sizes of 25–50 persons for fully mobile hunter-gatherers, somewhat higher for semi-sedentary groups, against indications of participation from a catchment of roughly 200 km. Notroff, Dietrich and Schmidt (2014) treat the building process itself as community-forming.

9. Monuments before farming: sequence and what it implies

The faunal evidence is consistent with hunting rather than herd management. Peters and Schmidt (2004) report that "remains of males rather than of females dominate in the samples," with adult-dominated profiles, the signature of hunted rather than managed populations. No domesticated animals are indicated. On the plant side, cereals are present in quantity, indicated by the ground-stone corpus and by phytolith analysis rather than by charred macro-remains, which are weakly preserved (Dietrich et al. 2019). The published work does not claim morphologically domesticated crops, and should not be read as doing so.

The regional context weakens the clean before-and-after that the popular formulation requires. Willcox and Stordeur (2012) document large-scale cereal processing in northern Syria during the tenth millennium cal BC, ahead of domestication. Willcox (2013) and Asouti and Fuller (2013) both argue for a protracted, contextual emergence of agriculture rather than a threshold event. If cultivation was a long process with no single crossing point, then "before farming" specifies a position on a gradient rather than a side of a line, and the rhetorical force of the phrase does not survive the translation.

The defensible sequence is therefore narrower than the slogan. Monumental construction at Göbekli Tepe begins before morphologically domesticated crops and managed animals appear in the region, but not before intensive, organised, large-scale plant management. That is still a substantive claim, and it still constrains any model in which monumentality requires an agricultural surplus. It does not license the inference that cult caused agriculture, and it does not settle whether the assembling groups were sedentary, mobile, or seasonally both. Peters et al. (2020) survey the domestication evidence for the site in the terms that evidence supports.

Schmidt's "first came the temple, then the city" (Schmidt 2000) is thus best handled as a proposal about causal direction whose antecedent has been complicated and whose consequent was never established. Banning (2011, 2023) contests it directly, and the protracted-emergence models make it difficult to state precisely. Schönicke (2018) places the site's abandonment at the end of the ninth millennium BC, alongside the slow advent of agriculture and animal husbandry, a coincidence in time compatible with several causal stories and selecting none of them. The slogan's afterlife has been considerably larger than its evidential base.

10. Pillar and megalith traditions: continuity, convergence and the limits of comparison

Regional comparanda are legitimate and instructive. The Pre-Pottery Neolithic A tower at Tell es-Sultan, Jericho, excavated by Kathleen Kenyon in the 1950s, is where Kinzel and Clare (2020) open their discussion of monumentality, and is the obvious near-contemporary case of collective construction without agricultural surplus. Its commonly cited dimensions (some 8.5 m high, about 9 m in basal diameter, an internal stair of twenty-two steps) could not be verified against Kenyon's primary publications for this review, and are reported as conventional figures. An astronomically framed reading has been proposed by Barkai and Liran (2011), who argue that the shadow of a nearby mountain first reaches the tower at midsummer sunset and treat it as a mechanism for motivating communal participation. It is published as a short project-gallery report, and it is the kind of argument the Göbekli excavators reject for their own site, an asymmetry better named than exploited.

Atlit-Yam, now submerged off the Carmel coast, contains a semicircle of seven megaliths of roughly 600 kg each, bearing cup marks and arranged around a freshwater spring, together with a dry-stone well 1.5 m in diameter and 5.5 m deep (Galili et al. 1993). The site is dated 8100–7900 BP on charcoal and seed samples, in the uncalibrated convention its excavators used. It is Pre-Pottery Neolithic C, roughly two millennia later than Göbekli's monumental enclosures, and it is a fishing village. As an instance of Levantine standing-stone practice it is apt; as evidence of a continuous pillar tradition it is not.

Nabta Playa sits further still from the case. Malville et al. (1998) describe a ceremonial complex in southern Egypt with alignments to cardinal and solstitial directions and five megalithic alignments radiating from possible funerary structures. Malville (2015) places the centre in the Late and Terminal Neolithic at 7,500–5,400 BP and writes that the megaliths "may once have aligned" with Arcturus, the belt of Orion, Sirius and α Centauri, with reorientations that "suggest" a response to precession. Claims of this kind attract criticism, and the status of that criticism should be stated rather than assumed: no published rebuttal specifically addressing the stellar alignments could be located for this review. The defensible formulation is not that they have been refuted in print but that they are hedged by their own proponents and remain independently unconfirmed. The complex postdates Göbekli's enclosures by three to five millennia.

That brings comparison to its limit. The suggestion that Göbekli Tepe's T-pillars stand in continuity with Andean megalithic masonry, Sacsayhuamán is the usual example, is not supportable at any level. The chronological gap is roughly ten and a half millennia, the geographic gap is a hemisphere, and no intermediate distribution, material continuity or transmission mechanism has been proposed. The technologies differ as well: fitted polygonal ashlar in a retaining wall solves a different problem from free-standing carved monoliths set in sockets. Convergent invention is the defensible reading, and it should be stated in those terms rather than left implied.

The provenance of the continuity idea is worth recording. It belongs to a popular literature that Banning (2023) addresses by name, alongside claims associating the site with a lost pre-Holocene civilisation or with extraterrestrial visitors; critical replies have appeared in Skeptic 22 (3), Scientific American 316 (6) and American Antiquity 80 (3). A better alternative exists: Watkins (2020) treats monumentality in Neolithic southwest Asia as a technology of memory in time and space, and Kinzel and Clare (2020) argue that monumental status is relational and perceptual rather than a function of absolute size, asking whether it derives "from physical size alone or from the concepts behind the building and how the responsible communities perceived them." Read that way, recurrence across unconnected traditions is interesting precisely because the traditions are unconnected.

11. Reading the site as residual expression

The reading proposed here is that Göbekli Tepe is best understood as residual expression: a set of practices whose monumental form is the visible remainder of something already under way, rather than a founding act from which later developments follow. The formulation is offered as a reading, not as a finding. Its warrant is coherence with the revised evidence, and its principal virtue is that it does not require the sequential architecture the excavation team has dismantled. It is also, deliberately, a weaker claim than the one it replaces: a relation between registers of activity rather than a cause.

That dismantling forces a specific correction to how the thesis must be stated. On Schmidt's scheme one could describe a decline from monumental round-oval enclosures to modest rectangular buildings, and read the latter as residue of the former. Clare (2020) removes the option: the two building types overlap, multi-phase structures were long-lived, and some final building phases are consistent with construction around the mid-ninth millennium cal BC. The residual and the monumental were partly contemporaneous. "Residual" therefore has to be read synchronically, as persisting alongside, rather than sequentially, as left over after.

In that synchronic sense the term earns its place. Domestic buildings, hearth sequences, bead production, cisterns, sub-floor burials and thousands of grinding tools occupy the same site and, in part, the same centuries as the great enclosures. The monumental is then not a phase but a register: an intensification, occasional and expensive, of practices otherwise conducted at ordinary scale. On this reading the enclosures are not what the community was for; they are what it could do, periodically, when it assembled. Özdoğan's (2024) formulation for Sayburç, the intertwining of the sacred with everyday life, describes the same relation from the other direction.

The reading also accommodates the abandonment evidence without requiring a ritual finale. If the mound formed largely through slope-derived inundation with prehistoric mitigation (Clare 2020) rather than through deliberate closure, the end of the site is not a symbolic statement but an accumulation of ordinary events (erosion, deposition, repair, eventual disuse) of the same kind as its middle. That is less satisfying narratively and more consistent evidentially. Kinzel and Clare's (2020) relational account of monumentality is congenial to it: if monumental status inheres in concepts and events rather than in mass, its residue need not take the form of a ruin.

What the reading does not do is explain why the pillars were made. It reorganises the relation between the monumental and the domestic; it says nothing decisive about the anthropomorphic figures, the animal reliefs, the treatment of skulls, or the mismatch between what was carved and what was consumed. Peters and Schmidt's 2004 judgement has not been superseded, and bears repeating: the question of what purpose the enclosures served will take much more time to be properly answered. A reading that respects that sentence is preferable to one that circumvents it.

12. What would count against this reading, and what remains open

The reading is falsifiable in several specific ways, which is the main argument for stating it in this form. A published Bayesian chronological model recovering a clean succession (monumental enclosures completed before rectangular buildings begin, with no significant overlap) would restore the sequential narrative and make the synchronic formulation unnecessary. Clare's (2020) overlap rests on radiocarbon dating of mud-mortar organic residues from a limited number of structures. A larger, properly modelled series could revise it in either direction, and until one is published the revision itself should be held as provisional.

Evidence of deliberate closure would matter equally. If sedimentological work on the fills recovered depositional signatures inconsistent with slope failure, or stratigraphic sequences indicating rapid single-episode filling of an enclosure, the earlier reading would be reinstated on better grounds than it originally had. Karul's (2021) argument for deliberate filling at Karahantepe shows the practice is live in the region, and Schönicke's continuing work on abandonment processes is where a Göbekli-specific test would come from. On the present state of publication, the sedimentology of the fills is the single most decisive outstanding question.

Three further outcomes would press on the reading. Confirmation that the large enclosures were unroofed would reopen the astronomical readings that roofing currently excludes (Kurapkat 2012; Notroff et al. 2017). Excavation of large pillars, in the 4–5.5 m class, at a second site would undercut the argument that Göbekli Tepe is an outlier in scale within its region, given the pace of the Taş Tepeler programme, the likeliest of the three. Identification of substantial storage facilities would remove the constraint on which the work-feast model depends (Dietrich et al. 2019), converting periodic assembly into something closer to sustained residence.

The open questions are correspondingly concrete: the content of the 2020 integrated trough study and what it establishes about fermentation; whether the geophysical estimate of at least twenty large enclosures survives systematic excavation; the absolute chronology of Nevalı Çori as verified from Hauptmann's reports; and the taphonomy of the fills in detail. Each is answerable, and none is answered. The appropriate posture towards a site of this importance and this publication history is to hold the reading loosely and to name the observations that would change it, coherence being the measure, not final demonstration.