Mesolithic dates at Stonehenge and the bluestone quarries

Trench 1 at Carn Goedog in 2016, viewed from the south (photograph by Adam Stanford). Pearson, Mike Parker, Josh Pollard, Colin Richards, Kate Welham, Chris Casswell, Charles French, and others, ‘Megalith Quarries for Stonehenge’s Bluestones’, Antiquity, 93 (2019), 45–62 http://dx.doi.org/10.15184/aqy.2018.111

Every so often someone rediscovers that Mesolithic radiocarbon dates keep appearing in the Stonehenge story — beneath the old car park, at Craig Rhos-y-Felin, at Carn Goedog — and presents this as an anomaly the discipline has declined to confront.

The dates are real. They are also all published, most of them in the same reports that establish the Neolithic chronology, and those reports are considerably more candid about the awkwardness than the rediscoverers tend to notice. What follows sets out what has actually been dated, at what depth, in what context, and on what material, because that is where the argument lives.

Two things emerged from doing this carefully that I had not expected. "Mesolithic" is doing far too much work as a single category: the Early Mesolithic and Late Mesolithic material at these sites occur in different kinds of context and mean different things. And the single find most often offered as a bridge between Stonehenge and the quarries — a rhyolite fragment in a car park pit — comes from a layer that was never radiocarbon dated.

Declaration: I am a co-author on Bevins et al. (2025), cited below, and have published separately on the glacial transport question.

The car park pits: what was actually found

Faith and Lance Vatcher excavated the car park extension between 7 February and 18 March 1966, ahead of works by the then Ministry of Public Buildings and Works (trench C82). Three circular holes appeared as the surface was cleaned down to chalk, in a line running roughly east–west on the 315ft contour, with a fourth, irregular disturbance further west. Centre to centre, A to B is 9.75m and B to C is 12.29m, with a further 13.72m from A to the irregular feature.

All three held large posts. The post in A was about 66cm across at 75cm depth, B about the same, C between 61 and 83cm. Holes A and B had recesses cut into the chalk on their south-east sides that the Vatchers read as seats for wedging timbers, and in B the section cut clean across traces of two such timbers running from the side of the hole to the side of the post. Both B and C had material accumulated on the base before the post went in — a 5cm scatter of soil in B, five thin layers in C — so the holes stood open for a while after being dug. Pit C did not show a clear postpipe; a post may have slipped out when the timber was largely rotten.

The Vatchers assumed a Late Neolithic date. What changed that was Susan Limbrey's charcoal identification: predominantly pine, a species not expected on the chalk in the Late Neolithic. That identification is what prompted the submission of two samples for dating.

One correction is due here. Allen states that all the material submitted to Limbrey was pine. Her own appendix to the Vatcher report says most of it was, and records that two samples from Hole C contained Rosaceae of "Crataegus type" — hawthorn, apple or whitebeam — alongside the pine. The pine dominates, but it is not the only species present, and arguments about deliberate selection of timber should be built on the appendix rather than the summary.

Allen's Table 3 in Stonehenge in its Landscape is worth reproducing, because it contains detail that rarely travels:

Feature	Diam (m)	Depth (m)	Flint	Rhyolite chip	Charcoal	Decayed wood	Burnt bone	Antler pick marks	Lab no	Determination	Calibration
Treehole	2.5	0.71	–	–	–	–	–	–	–	–	–
Pit A	1.93	1.34	–	–	Yes	Possibly	–	Yes	HAR-455	9130±180 BP	8820–7730 cal BC
Pit B	1.27–1.47	1.27	–	–	Yes	Yes	Yes (1)	Yes	HAR-456	8090±140 BP	7480–6590 cal BC
Pit C	1.52–1.77	1.55	Yes	–	Yes	Yes	–	Yes	–	–	–
Pit 9580	1.9	1.3	Yes (4)	Yes (1)	Yes	–	–	–	see below

Three points follow that the popular summaries lose.

Only two of the Vatcher pits were dated. Pit C has no determination at all. The commonly quoted range for the car park posts — "8500 to 7000 BC", or Allen's own summary phrasing of 8500–7650 cal BC — characterises the complex; it is not a set of measurements. The two determinations calibrate to 8820–7730 and 7480–6590 cal BC. HAR-456's range falls entirely outside Allen's summary range, and both carry large errors (±180 and ±140 BP) typical of their vintage. Anyone building an argument on tight chronological coincidence between these posts and something else should look at those error terms first.

All three postpits show antler pick marks. The Vatchers put it as "apparently dug out with antler picks", so this is an inference from tool marks rather than a recovered tool. It is not a chronological argument either — antler picks are ubiquitous in the Neolithic — but it does establish that these were deliberately excavated features rather than natural hollows.

The relationship to Stonehenge is undetermined, and Allen says so. He notes that the pits' span is too great for them to have supported horizontal members, that they are more likely individual uprights, and that the line of poles may represent formal display of some kind predating Stonehenge by over four millennia. Whether the five features were contemporaneous cannot be confirmed on radiocarbon grounds — his words, not a sceptic's.

It is worth registering that the 1973 report predates the radiocarbon dates entirely. The Vatchers had no pottery and no absolute dating, judged the holes Neolithic on morphology, and closed by endorsing C.A. Newham's argument that the three lined up with solar and lunar setting phenomena viewed from the Four Stations and the Heel Stone — which would have required posts standing some 30ft above ground. The Mesolithic determinations arrived afterwards and removed the frame the excavators had published in. That is worth remembering when the car park posts are pressed into service as evidence of continuity with the monument: the one published attempt to connect them to Stonehenge directly was made on the assumption that they were four thousand years younger than they are.

Pit 9580, and the rhyolite fragment

This find carries almost the whole weight of the Mesolithic-Stonehenge argument, so it deserves the full sequence.

Pit 9580 was recorded by Wessex Archaeology in 1988–9 (Martin Trott's unpublished notes, synthesised by Allen). The original feature was a circular hole cut into soft weathered chalk, about 1m across and 1.3m deep, filled with horizontally bedded layers of clean chalk rubble (contexts 9587–9592). Allen reads the horizontal bedding and clean chalky character as deliberate backfilling rather than natural silting.

At a later date the pit was widened — a probable recut — producing a broad shallow pit about 1.9m across and only 0.7m deep, with its own primary (9588), secondary (9585) and tertiary (9581–2) fills. Allen's reading is that the feature was originally a postpit like the others, and was subsequently partially backfilled, altered and redesigned, possibly because a post was removed and the hole had to be enlarged to free it.

The determinations, from Allen's Table 4, are all on Pinus charcoal:

Fill	Context	Depth (m)	Lab ref	Determination	Calibration
Tertiary	9582	0.25–0.35	OxA-4220	8400±100 BP	7580–7090 cal BC
'Recut': secondary	9585	c. 0.40	OxA-4219	8520±80 BP	7700–7420 cal BC
'Recut': base of secondary	9585	0.57	GU-5109	8880±80 BP	8090–7690 cal BC

Context 9581 does not appear. The soil monolith sampled the secondary fill and the tertiary sequence, but Allen states that samples of the upper tertiary fill — 9581 — were not taken. The rhyolite fragment, 62g, sat at 0.2m depth in 9581.

That disposes of the strongest version of the sceptical reading. Brian John has argued that the rhyolite must be Mesolithic because a Mesolithic determination came from the tertiary fill, and that the layer is essentially homogeneous with no stratigraphic break to justify a later intrusion. But 8400±100 BP is from 9582, a different context 5–15cm deeper, and 9581 is undated. The date and the find are not in the same layer.

Allen's own reasoning about the fragment is also partly circular, and he does not quite acknowledge it. He treats the rhyolite as indicating that 9581 is contemporary with the dressing of the bluestones, then cites 9581's phase 3 attribution when discussing the environmental sequence. Scaife does the same in his pollen report within the same chapter. If that were the whole argument it would not be worth much.

It is not the whole argument. The molluscan evidence stands independently:

• Local mollusc zone 1 (chalk backfill and primary recut fills): dominated by shade-loving species, the Zonitids and the Punctum group, with Ena montana and Euconulus fulvus indicating mature ancient woodland with some leaf litter.
• Zone 2 (recut secondary fill 9585): gradual increase in open-country species, decline in E. fulvus, absence of E. montana, Pupilla muscorum and Helicella itala present. Open woodland with grass.
• Zone 3 (tertiary fills 9581, 9582): Allen describes a distinct break rather than a transition. Rapid increase in open country species, marked decline in shade-loving elements, P. muscorum superabundant at 82%, low overall diversity.

The decisive independent detail is that this Pupilla superabundance is mirrored in Evans's ditch zone F at Stonehenge, which is stratigraphically phase 3 within the Stonehenge Ditch. That correlation owes nothing to the rhyolite. Scaife's pollen sequence shows the same break at 360mm, between PAZ1 and PAZ2. Allen concludes that there is a hiatus of about five millennia in the pit's environmental record — Boreal woodland below, Sub-boreal open grassland above, the Atlantic simply unrepresented.

So the chip comes from an undated context, at the top of a reworked pit, on the far side of a break independently attested by snails and pollen and independently correlated to the Stonehenge Ditch. It sits 0.2m below the modern surface, in an area where W.E.V. Young recovered further pieces of both bluestone and sarsen, including several mauls, from the car park in 1935.

Why the fragment has been set aside

The fragment has never, as far as I can establish, been characterised petrographically. It does not appear in the provenancing literature. My reading is that it has been treated as intrusive and not diagnostic, and passed over on that basis.

That judgement looks sound, and the distribution across the car park features supports it. Of the five, only pit 9580 produced any bluestone. The three postpits whose timbers rotted in situ, and the treehole, produced none — and those are the features that were dug once and left. Pit 9580 is the one that was dug, deliberately backfilled, widened, and refilled, with its uppermost fill 0.2m below a surface that has since been ploughed, landscaped, tarmacked and stripped. The bluestone appears in the disturbed feature and nowhere else.

There is a useful control on how far small bluestone debris travels in such deposits: when Aubrey Hole 7 was re-excavated in 2008 it produced 63 fragments of bluestone and 455 of sarsen, despite having already been excavated and backfilled twice, and detailed analysis showed all the bluestone to be chippings rather than clasts. A single 62g chip near the top of a recut pit in the Stonehenge landscape is an unremarkable object.

But it is worth being honest about the shape of the reasoning, because it is the same shape I have just criticised in Allen. The standing position in the literature is that no bluestone debris from the Stonehenge landscape has been found in a securely pre-3000 BCE archaeological or natural context. In this one instance that claim rests partly on the assumption that the fragment is intrusive, rather than on an analysis that establishes it. The assumption is well founded, and the stratigraphy carries the argument without any help from the stone — but it is an assumption.

It could be closed. Bevins et al. (2025) set out the diagnostic criteria for Craig Rhos-y-Felin rhyolite: needle-like stilpnomelane crystals lying along and overlying a strong foliation, with secondary chlorite and titanite and partially resorbed primary zircon — the only documented occurrence of stilpnomelane in a foliated rhyolite anywhere in Wales, and not found elsewhere in north Pembrokeshire despite examination of hundreds of samples. Geochemically, Craig Rhos-y-Felin rhyolites all carry less than 500 ppm Zr and a Zr/Th ratio of about 20, against ratios above 40 for many other north Pembrokeshire rhyolites. Every piece of foliated rhyolite debitage so far analysed from Stonehenge — the Newall boulder, Cunnington's 1880s collection, the Open University samples used by Thorpe et al. — plots inside that field.

Portable XRF is non-destructive. If the fragment survives in an accessible archive, an afternoon's work would convert an assumption into a measurement. There is a prior question too, and a cheaper one: the identification as "rhyolite" comes from a field record in unpublished notes, and has never been checked by a petrographer. Whether the object is rhyolite at all is the first thing anyone should establish.

None of which changes the conclusion. It would simply put it on a different footing.

Craig Rhos-y-Felin: the sequence

Rhosyfelin has about 2m of stratigraphy on the west side of the outcrop, running from the Early Mesolithic to the medieval period on a bed of glacial till.

Four Early Mesolithic hearths were set sequentially into the top of a large dug pit. Eight determinations on carbonised hazelnut shell and roundwood:

Context	Date cal BC	Date BP	Sample	Material
111	8550–8330	9229±21	SUERC-50761	Nutshell, Corylus
106	8530–8280	9157±40	OxA-30507	Nutshell, Corylus
100	8290–7970	8984±47	SUERC-51164	Roundwood, Corylus
102	8290–7970	8970±45	OxA-30548	Roundwood, Corylus
102	8240–7850	8890±40	OxA-30549	Roundwood, Corylus
111	8210–7960	8888±21	SUERC-50760	Roundwood, Corylus
106	8210–7790	8851±44	SUERC-51165	Roundwood, Corylus
100	8210–7790	8848±37	OxA-30506	Roundwood, Corylus

Note the sample selection: hazelnut shell and roundwood throughout, both short-lived, minimising the old-wood offset. This is deliberate and it matters. When the same project dates oak charcoal — buried soil 098, at 7540–7300 and 7460–7180 cal BC — the material could be centuries older than its context of deposition. Any argument leaning on precise coincidence between determinations needs to know which kind of sample it is leaning on.

The published verdict on Mesolithic quarrying is short and unambiguous: the only artefact from the entire hearth sequence was a single tiny flint flake, and there was no evidence of any Mesolithic quarrying or working of rhyolite at the outcrop.

The quarry-associated features sit in the Neolithic occupation layer above (059 = 159), dated by two hazelnut shells to 3620–3360 and 3500–3120 cal BC. The prone monolith and its platform are Early Bronze Age, later than 2140–1950 cal BC.

Early Mesolithic, Late Mesolithic: an important distinction

Both quarry reports use "Mesolithic" as a single label, and this obscures something. Sorting the determinations by period changes the picture.

Early Mesolithic material (9th–7th millennia BC) at the quarries occurs in four places: Rhosyfelin's hearths and buried soil; Rhosyfelin's two orthostat-setting pits (110 and 116: 7940–7650, 8280–7970, 8190–7680 cal BC); Carn Goedog's central hearth (7190–6840 cal BC, OxA-31823); and two Carn Goedog platform-related contexts (176 at 7590–7380, OxA-35184; 130 at 6760–6530, OxA-35157).

Late Mesolithic and transitional material (6th–5th millennia BC) occurs somewhere quite different — overwhelmingly in redeposited and alluvial contexts:

Site	Context	Date cal BC	Date BP	Sample
Carn Goedog	171, middle ditch fill (east)	5470–5230	6359±33	OxA-35156
Carn Goedog	171, middle ditch fill (east)	4910–4690	5910±45	OxA-35396
Carn Goedog	128, upper ditch fill (west)	4530–4360	5619±34	OxA-35153
Carn Goedog	135, middle ditch fill (west)	4450–4330	5521±34	OxA-35155
Carn Goedog	128, upper ditch fill (west)	4230–3960	5236±34	OxA-35395
Rhosyfelin	164, platform fill	5226–5011	6182±35	OxA-35149
Rhosyfelin	164, platform fill	4907–4723	5940±33	OxA-35150
Rhosyfelin	153, palaeochannel basal fill	5800–5640	6833±40	OxA-32021
Rhosyfelin	153, palaeochannel basal fill	5620–5460	6543±37	OxA-32022
Rhosyfelin	069	5210–4950	6114±31	SUERC-46204

This matters for two reasons.

First, Parker Pearson et al. describe the ditch-129 dates as Mesolithic in their running text. They are — but Late Mesolithic, not the Early Mesolithic of the central hearth. Anyone repeating the phrase "Mesolithic hearths at both quarries with the same chronology" is conflating determinations two thousand years apart.

Second, the ditch dates make the residuality reading close to inevitable. Ditch 129 was cut through a podzol and backfilled with rubble and the iron-rich lower B horizon material dug out to make it. Charcoal in redeposited soil dates the soil's accumulated carbon, not the digging. The one determination in that ditch that does date the event — 3020–2880 cal BC, OxA-35154, from the middle fill — is Neolithic and matches the platform.

Paired dates: the residuality signature

The cleanest way to see the problem is to look at contexts that produced more than one determination:

• Carn Goedog central hearth (106): 7190–6840 cal BC and 2890–2630 cal BC. Four millennia apart, from one hearth set in a gap where a slab had been lifted out of the platform.
• Carn Goedog platform sediment (176): 7590–7380 cal BC and 3940–3690 cal BC.
• Carn Goedog sediment above platform (130): 6760–6530 cal BC and cal AD 1520–modern.
• Carn Goedog pillar recess (119): 2130–1900 cal BC and cal AD 1680–1940.
• Rhosyfelin Early Bronze Age platform (115): 2140–1950 and 2200–1980 cal BC and 4330–4050 cal BC.
• Rhosyfelin Iron Age pit (047): 3095–2925 and 2840–2495 cal BC — Neolithic hazelnut shell in an Iron Age feature.

This is what a landscape with long occupation and a lot of soil movement produces. The chronological model published for Rhosyfelin (Hamilton, in Parker Pearson et al. 2015, fig. 7) colour-codes the determinations inconsistent with the stratigraphy, separating suspected later contaminants from suspected residual material. The disagreements are the subject of the figure, not something concealed beneath it.

Note the direction the argument runs. A "take every date at face value" rule applied consistently would put megalith erection at Rhosyfelin in the Early Mesolithic and Neolithic activity in an Iron Age pit and early modern charcoal in a Bronze Age recess. Applied selectively, it produces whichever period the author started with.

Where the argument still has soft ground

Two things should be conceded rather than argued around.

The orthostat pits at Rhosyfelin are the weak joint. One of the two orthostats interpreted as a fulcrum for moving monoliths was set into a larger Early Mesolithic pit containing the sequence of hearths. John, Elis-Gruffydd and Downes have pressed exactly here, objecting that the stratigraphic separation is asserted rather than demonstrated. They are entitled to press. The residuality reading for contexts 110 and 116 is an interpretation, and its warrant is the wider stratigraphic model rather than a directly observed cut. The Neolithic case at Rhosyfelin does not depend on these two features — it rests on the dated occupation layer beside the recess, and on two stone wedges found in situ within the joints around a pillar next to that recess — but the orthostat pits should not be cited as though they were secure.

Some of the 2019 language outran its data. The claim in that paper that the quarry evidence conclusively invalidates glacial transport was stronger than what the excavation alone could show. The case against glacial transport has since been made properly and on other grounds — provenancing of the Newall boulder, the sarsen sourcing, the absence of glacial deposits or of spotted dolerite erratics east of Narberth, the Altar Stone's Orcadian source — but a quarry excavation does not by itself settle a question about ice.

What should be resisted is the idea that these soft spots help an early chronology. They do not. The glacial argument and the Mesolithic-quarrying argument are incompatible: one holds that nobody quarried anything, the other requires that somebody did, four thousand years earlier than the excavators say. Neither supports the other.

What would count as evidence

The Neolithic case at Carn Goedog does not rest on charcoal. It rests on recesses in the rock face with fresh surfaces where pillars are absent and no trace of those pillars in the rubble below; an artificial platform of split slabs pressed into the underlying sediment by something heavy; fifteen wedge-profiled coarse stone tools of imported mudstone and sandstone with flake scars and battered terminals; and two recorded instances of joint-widening on the outcrop itself. At Rhosyfelin it includes wedges left in situ in the joints beside the extraction niche. The charcoal dates the soils these things sit on and in.

An equivalent Mesolithic case would need equivalent things — extraction features stratified within or beneath the Early Mesolithic horizons, quarrying tools in Mesolithic contexts, dolerite or rhyolite working debris in the hearth layers. At Rhosyfelin the hearth sequence produced one small flint flake. That absence is the argument, and it is a strong one.

The Atlantic gap, and Blick Mead

One further thing falls out of reading Allen properly. Writing in 1995, he observed that nowhere in the pit 9580 sequence is the Atlantic — later Mesolithic — represented, and that there was then a complete absence of datable late Mesolithic activity in the Stonehenge area. He was careful to add that this might reflect the pattern of fieldwork rather than past human behaviour.

He was right to hedge. Blick Mead, at the spring below Vespasian's Camp about 2km east of Stonehenge, has since produced tens of thousands of struck flints, an aurochs-dominated faunal assemblage, and a radiocarbon sequence running from the eighth millennium down to roughly 4200–4000 BC — among the latest Mesolithic determinations from anywhere in England. The gap Allen identified has been substantially filled by a site nobody had excavated when he wrote.

This is the useful frame for all of it. Mesolithic material keeps appearing near Stonehenge and at Preseli crags because these were long-used Mesolithic landscapes. Sheltered outcrops beside water and open downland beside springs are exactly where that use is visible. Finding traces of it under a place later monumentalised is the expected result, not the anomaly.

The short version

At the quarries, Early Mesolithic material lies stratigraphically below the extraction features or appears as residual charcoal in disturbed contexts, and the Late Mesolithic material appears almost exclusively in redeposited ditch fills and alluvium. At Stonehenge, the Mesolithic features are 200m from the monument and four millennia earlier than anything proposed for it. The one find offered as a bridge between the two comes from an undated layer above a five-millennium environmental hiatus, in the top of a pit that had been dug, backfilled, widened and refilled.

A radiocarbon determination tells you when an organism stopped exchanging carbon. Establishing what it dates — which event, which deposit, which act — is a separate exercise, and it is most of the discipline.

---

Sources

• Allen, M.J. 1995. Before Stonehenge, in R.M.J. Cleal, K.E. Walker & R. Montague, Stonehenge in its Landscape: twentieth-century excavations (English Heritage Archaeological Report 10): 41–62. Includes Scaife's pollen report and Table 4.
• Vatcher, L. & Vatcher, F. 1973. Excavation of three post-holes in Stonehenge car park. WANHM 68 (Part B: Archaeology and Local History): 57–63. Includes Limbrey's charcoal identification appendix. Scanned at the Biodiversity Heritage Library
• Parker Pearson, M. et al. 2015. Craig Rhos-y-felin: a Welsh bluestone megalith quarry for Stonehenge. Antiquity 89: 1331–52. Open access: doi.org/10.15184/aqy.2015.177
• Parker Pearson, M. et al. 2019. Megalith quarries for Stonehenge's bluestones. Antiquity 93: 45–62.
• Parker Pearson, M. et al. 2022. Reconstructing extraction techniques at Stonehenge's bluestone megalith quarries in the Preseli hills of west Wales. JAS: Reports 46: 103697.
• Bevins, R.E., Pearce, N.J.G., Ixer, R.A., Scourse, J., Daw, T., Parker Pearson, M., Pitts, M., Field, D., Pirrie, D., Saunders, I. & Power, M. 2025. The enigmatic 'Newall boulder' excavated at Stonehenge in 1924: new data and correcting the record. JAS: Reports 66: 105303. Open access: doi.org/10.1016/j.jasrep.2025.105303
• Bevins, R.E. et al. 2023. Lithological description and provenancing of a collection of bluestones from excavations at Stonehenge by William Hawley in 1924. Geoarchaeology 38: 771–85.
• Darvill, T., Marshall, P., Parker Pearson, M. & Wainwright, G. 2012. Stonehenge remodelled. Antiquity 86: 1021–40.
• Jacques, D. & Phillips, T. 2014. Mesolithic settlement near Stonehenge: excavations at Blick Mead, Vespasian's Camp, Amesbury. WANHM 107: 7–27.
• Ixer, R.A. & Bevins, R.E. 2011. Craig Rhos-y-felin, Pont Saeson is the dominant source of the Stonehenge rhyolitic 'debitage'. Archaeology in Wales 50: 21–31.
• John, B.S., Elis-Gruffydd, D. & Downes, J. 2015a. Quaternary events at Craig Rhosyfelin, Pembrokeshire. Quaternary Newsletter 137: 16–32.
• John, B.S., Elis-Gruffydd, D. & Downes, J. 2015b. Observations on the supposed 'Neolithic bluestone quarry' at Craig Rhosyfelin, Pembrokeshire. Archaeology in Wales 54: 139–48.