How rock art is dated
A painting carries no signature and no date. Everything this atlas says about age rests on one of nine methods — some counting atoms, some reading the order of layers, one frankly admitting it is a comparison. Here is what each of them does, and where each of them stops.
Radiocarbon (AMS)
Everything alive takes up carbon from the air, including a trace of the unstable isotope carbon-14. When the organism dies the intake stops and the clock starts: the carbon-14 decays at a known rate while the stable carbon stays put. Measure the ratio and you have the time since death. Accelerator mass spectrometry counts those atoms individually rather than waiting to detect their decay, which is why a sample the size of a poppy seed is now enough where a matchbox of material was once required.
Rock art can be dated this way only where the paint itself is organic — charcoal above all, which is why black figures are so often the ones with a number attached and red ochre ones so rarely are. Ochre is a mineral and contains no carbon to count. Where the pigment cannot be sampled, the method is turned on something near it instead: a torch mark on the wall, charcoal from a hearth in the floor, a bone in the layer that buried the panel.
The clock times the death of the organism, not the making of the image. Charcoal from a long-dead tree can be centuries older than the hand that drew with it, and a hearth below a painting only proves people were there, not that they painted that day. Contamination pushes results in both directions, and beyond about fifty thousand years so little carbon-14 survives that the method fades out altogether — which is precisely the range where the oldest art lies.
Chauvet-Pont d'Arc · Font-de-Gaume · Gargas Cave · Les Combarelles · Cueva de Tito Bustillo · Cave of Nerja · and 13 more
Uranium-series on calcite
Water seeping through limestone carries dissolved uranium and lays it down in the thin calcite crusts that grow over cave walls — the same films that make flowstone and cave popcorn. Uranium decays into thorium at a known rate, and thorium is not soluble, so none of it arrives with the water: every atom present grew there. The ratio of the two dates the crust.
The crust is dated, never the painting. A film that formed over a painted surface must be younger than the paint beneath it, so its age is a minimum: the art is at least that old. A crust the paint was applied on top of gives a maximum. Sampling both, above and below, brackets the image between two dates. Laser ablation now takes those samples in slivers thin enough to leave the art untouched, which is what opened the method to panels no one would previously allow a drill near.
It brackets rather than pins. A minimum age of forty thousand years is compatible with a painting made forty thousand years ago or two hundred thousand, and only the upper bracket closes that gap. The method also assumes the crust behaved as a sealed system — that no uranium washed out and no thorium washed in — and where that assumption has been challenged, so have the dates. Several of the most striking results in this atlas are argued over on exactly those grounds.
Altamira Cave · Cueva de El Castillo · Cueva de Ardales · Leang Tedongnge · Maros-Pangkep · Creswell Crags (Church Hole) · and 5 more
Optically stimulated luminescence
Buried quartz and feldspar grains sit in a faint bath of natural radiation from the surrounding sediment, and the energy knocks electrons into flaws in the crystal where they stay trapped. Sunlight empties those traps in minutes. So a grain that has been in the dark accumulates a charge in proportion to how long it has been buried, and shining a controlled light on it in the laboratory releases that charge as a measurable glow. The brightness is a clock reading the last time the grain saw daylight.
For rock art the grains are almost never in the art. They are in the sediment that buried a decorated panel, in the fill behind a fallen slab, or beneath a boulder that rolled across a carving — each of which fixes a moment the art must already have existed by. The method is at its best where an image is sealed by a datable event rather than exposed on an open surface.
The result dates the burial, and the distance between that and the making of the art can be anything from a season to millennia. It also depends on knowing the radiation dose the grains received, which means measuring the sediment around them and assuming it stayed constant — moisture changes alter the rate. Incomplete bleaching, where a grain was buried without a proper dose of sunlight first, leaves an inherited charge and makes a sample look older than it is.
Gwion Gwion (Kimberley) · Tassili n'Ajjer · Barrier Canyon (Horseshoe Canyon) · Uffington White Horse
Stratigraphic context
Deposits accumulate in order, oldest at the bottom, and archaeology reads them like the pages of a book. Anything found within a layer is contemporary with it or older; anything sealed beneath one is older still. The layers themselves are dated by whatever they contain that a laboratory can measure, and everything else in them inherits that bracket.
Art enters this sequence in two ways. A decorated block can fall from a ceiling and be buried, sealing it under everything deposited afterwards. Or the deposit can rise against a wall and cover the lower edge of a panel, in which case the art is older than the layer touching it. The most persuasive cases are the ones where a painted or engraved fragment is excavated from a dated layer rather than argued about on the wall.
This dates the relationship, not the image. A panel buried by a layer might have been made a week or ten thousand years before the burial, and the sequence gives no way to choose. Layers are also disturbed more often than they look — by burrowing animals, by later digging, by the floor being trampled for millennia — and a single artefact in the wrong place can drag a whole argument with it.
Chauvet-Pont d'Arc · Bhimbetka Rock Shelters · Tsodilo Hills · Cueva de las Manos · Abri Pataud · Castel Merle (Abri Reverdit) · and 20 more
Associated artefacts
Making images leaves debris. Ground pigment and the stones it was ground on, the stained tubes and bone pipes used to blow it, lamps that lit the wall, engraving burins, and portable objects carrying the same motifs as the panel above them. When such things are found in a dated deposit, they carry a date the art can be argued into.
The strongest form is a physical join: a fragment of engraved wall found in a layer, or a pigment crayon whose composition matches the paint on the panel. Weaker but common is the motif match — an engraved plaquette in a dated level bearing the same animal, drawn the same way, as the wall nearby. That is how much of Palaeolithic art is placed within a cultural period without a single direct measurement.
Association is an argument, not a measurement. A lamp on the floor of a painted cave shows someone brought light there; it does not show they painted. Objects move — down cracks, up through trampled floors, into the wrong box during a nineteenth-century excavation — and the closer the argument comes to resting on one find, the more it depends on that find being where the excavator thought it was.
Lascaux Cave · Abri de Cap Blanc · Isturitz & Oxocelhaya · Roc-aux-Sorciers · Roc-de-Sers · Nine Mile Canyon · and 6 more
Superimposition sequence
Where one image is drawn across another, the order is not in doubt: the covering image came second. Enough overlaps on enough panels build a relative sequence for a whole site, and where styles recur across a region that sequence can be extended between sites. It is the oldest technique in the study of rock art and needs no laboratory at all.
The reading is done at the surface, tracing which line runs over which and where a pigment sits on top of another's crust. Whole traditions have been ordered this way — a naturalistic phase overlain by a later schematic one, a red horizon beneath a white one — and the phases so defined become the framework everything else at the site is described against.
It yields order, never age. A sequence of five phases is equally compatible with five centuries or fifty thousand years, and gives no way to distinguish them without an absolute method attached at some point in the chain. It can also mislead: a later artist may have deliberately incorporated an older figure rather than ignored it, and weathering can make a faded upper layer look like the lower one.
Tassili n'Ajjer · Kondoa Rock Art Sites · Valcamonica Rock Art
Shoreline displacement
The last ice sheets were heavy enough to press the crust beneath them down, and since they melted the land has been rising slowly back — still rising today across Scandinavia and around the Baltic. The rate is known from cores, raised beaches and dated shorelines, so for those coasts there is a curve relating height above present sea level to the age at which that height was the waterline.
Carvings on those coasts were made at the water's edge, on rock the sea had polished and the tide kept clear. As the land rose the panels were lifted clear of the water, and a panel's height above the modern shore reads off the curve as the date it was still at the waterline. Where the carvings of a site spread across a range of heights, the sequence in height is a sequence in time.
Everything depends on the premise that the art was made at the shore — plausible where carvings cluster at old waterlines and worthless where they do not. The uplift curve is regional and imprecise, so ages come with wide error bars, and the method exists only where post-glacial rebound happened at all. Outside northern Europe it is unavailable.
Historical record
Sometimes the art dates itself against events that were written down. A depicted object that entered a region at a known moment — a horse, a firearm, a sailing ship, a wheeled cart — puts everything showing it after that moment. Inscriptions in a datable script do the same directly, and where a traveller described a panel in a dated account, the art is older than the description.
This works best on the recent end of the record, where rock art traditions ran on into periods with documents. Panels showing riders and guns are placed after contact; scenes of boats can be tied to known vessel types; inscriptions carry the date of the writing system they use. It also works negatively and usefully: art at a site that shows none of these things is likely to predate their arrival.
It dates a motif's earliest possible appearance, not the panel — a horse could be carved a year after the first horse arrived or three centuries later. Identification is often the weak link, since a schematic quadruped need not be a horse at all. And the method has nothing to say about the deep past, which is most of what this atlas contains.
Writing-on-Stone (Áísínai'pi) · Naj Tunich · Bir Hima (Ḥimā)
Stylistic comparison
Images made in the same tradition resemble one another — in what is shown, in how a body is outlined, in which conventions are used for horns, hooves and perspective. Where some examples of a tradition have been dated by other means, the resemblance is used to place the rest. This is the framework within which most rock art in the world is described.
A panel is compared with dated assemblages, above all with portable art from excavated levels, where the same conventions can be tied to a period. The result is an attribution to a culture or phase rather than a number of years, which is why so many sites in this atlas carry a period name where a date would be more satisfying.
It is the weakest form of evidence here, and the atlas treats it as the honest default rather than a result: it appears wherever nothing stronger is available. Styles are not clocks. They persist for millennia, revive after long gaps, and appear independently in places that never met, so a confident stylistic date can be wrong by any margin at all. Every absolute method that has been applied to rock art has moved at least one stylistic estimate, and usually in the direction nobody expected.
Lascaux Cave · Altamira Cave · Font-de-Gaume · Pech Merle · Niaux Cave · Cosquer Cave · and 136 more