Radiocarbon Dating: What an Archaeological Date Actually Means

What the instrument measures

Living organisms exchange carbon with their environment. After exchange stops, carbon-14 decays while stable carbon remains. Laboratories measure isotope ratios and compare them with standards and backgrounds. The result concerns the sampled carbon fraction, so material identification and chemical isolation are part of the measurement.

BP is a conventional radiocarbon scale

A laboratory age such as 2,000 plus or minus 30 BP is not simply 50 BCE. BP is defined relative to 1950 and uses conventional assumptions, including the Libby half-life. The quoted uncertainty describes measurement statistics under the laboratory model; it does not include every archaeological uncertainty.

Calibration converts radiocarbon age to calendar probability

Past atmospheric carbon-14 concentration varied. Independently dated archives, especially tree rings, establish calibration curves. A measured likelihood is mapped through the curve to calendar time, producing a non-uniform distribution rather than subtracting a number from 1950.

Choose the correct calibration curve

IntCal20 is designed for Northern Hemisphere atmospheric samples, SHCal20 for Southern Hemisphere atmospheric samples and Marine20 for non-polar marine material. Tropical convergence and mixed diets can require careful curve choice or mixing models. The curve and version must be reported.

A calibrated result may have several ranges

Wiggles and plateaus in the curve can map one measurement to multiple calendar intervals or a broad span. Report probability ranges and their masses, conventionally at 68.3 and 95.4 percent, rather than selecting the most convenient midpoint. Precision of measurement cannot remove a calibration plateau.

Sample-event association is the central question

Archaeologists usually want the date of construction, use, destruction, deposition or burial. The laboratory dates a sample. A valid inference requires a documented relationship between the organism's death, the sample's deposition and the event of interest.

Prefer short-lived, identified material

Seeds, annual plant remains, articulated bone or material formed shortly before deposition can reduce inbuilt age when context is secure. Taxonomic identification matters because a fragment of long-lived timber may predate felling and reuse by centuries.

Old wood and residual material

A tree's inner rings formed long before it was felled; deadwood can be stored or reused. Charcoal can also be eroded from an older deposit into a later feature. Such dates may provide a terminus after which deposition occurred, but should not automatically date the event.

Intrusion makes samples too young

Roots, burrowing animals, cracks and later pits can introduce younger carbon into older layers. Microstratigraphic observation, feature boundaries and sample integrity are therefore essential. A date inconsistent with stratigraphy should trigger investigation, not silent deletion.

Pretreatment removes unwanted carbon

Charcoal, wood, bone collagen, shell and residues require different chemical protocols. Laboratories record pretreatment codes, yields and quality measures such as collagen composition. Contamination is most consequential near the method's limits, and pretreatment can itself introduce offsets if poorly controlled.

Marine and freshwater reservoir effects

Marine organisms can appear older than contemporaneous terrestrial organisms because ocean carbon is not in immediate equilibrium with the atmosphere. Marine20 supplies a global baseline, often adjusted by a regional delta-R value. Freshwater hard-water effects and mixed marine diets can also bias dates.

The modern period has its own complications

Fossil-fuel emissions dilute atmospheric carbon-14, while nuclear-weapons testing sharply increased it after the 1950s. Bomb-pulse calibration can date recent tissues and materials, but multiple possible formation years and changing regional curves require explicit modelling.

One date rarely establishes a phase

Replicate or multiple samples help identify residuality, intrusion and laboratory problems. More dates are not automatically better: a defensible strategy targets questions, contexts and material likely to date the events of interest, with sampling designed before excavation or analysis ends.

Bayesian modelling combines dates and prior information

Programs such as OxCal can combine calibrated likelihoods with independently justified order, phase and boundary constraints. A posterior estimate may be narrower because the model adds information. Stratigraphy must be encoded honestly; the model cannot rescue incorrect associations.

Outliers belong in the model and report

A determination may disagree because of sample history, contamination, misidentification or an incorrect archaeological model. Formal outlier models and sensitivity analyses are preferable to removing inconvenient results. Publish exclusions and test how conclusions change.

Date the claim at the correct scale

A sample can support a context date; several contexts can define a phase; multiple sites contribute to a regional chronology. None alone dates an entire language, people or civilisation. The inferential distance should grow only with converging evidence.

Minimum reporting standard

Publish laboratory code, material and taxon, context, pretreatment, conventional age and uncertainty, carbon-isotope correction where relevant, calibration curve and software version, probability ranges, reservoir correction, model code, agreement diagnostics and reasons for exclusions.

How to read a headline date

Ask what was dated, where it came from, whether the context was sealed, whether the material was short-lived, which curve and reservoir correction were used, whether the range is calibrated, how many samples agree and whether the paper separates sample age from the historical claim.