How are archaeological finds dated?

The explanations distinguish finds, interpretations and open questions. Sources appear at the end of the page.

Before the number: date what?

Begin with a simple sentence: “I want to know when ... happened.” Complete it with an event: making an object, building a wall, using a floor or filling a pit. Then check whether the reported date concerns that event. A mismatch between the question and the measurement is a useful starting point, even without laboratory training.

Comparison and context

White, Paynter and Brown’s pottery guidance discusses material, form and technology alongside stratigraphic context. Source: M03. As a reader, ask what the find was compared with and how the comparison itself was dated. A useful comparison should lead you to the basis of the chronology, not just a similar-looking illustration.

What does radiocarbon measure?

The method measures carbon in a suitable sample of once-living material. A laboratory age is not a calendar year: calibration and reporting of ranges and probabilities are needed. Contamination control is central to measurement. Sources: M01, M04. When taking notes, retain the range and probability instead of copying only a convenient central number.

Combining evidence does not erase uncertainty

Bayesian chronological modelling combines measurements with other information, such as stratigraphic order, and depends on model assumptions. Source: M02. Ask which assumptions entered the calculation and how the model was checked against the data. A narrower result is not a reason to skip the methods.

A reading worksheet

• Event: what exactly is being dated?
• Evidence: what was measured or compared, and where did it come from?
• Context: how is the sample connected to the event?
• Result: which range and probability are reported?
• Limit: which assumptions or questions remain open?

Use the worksheet even if an abstract does not answer every question. Write “unknown from the available text”; there is no need to supply missing numbers to create an appearance of precision. This guide supports reading research. It is not a dating service or a method for authenticating an object from a photograph.

Example: three dates for one room

This example is invented for explanation. A room contains a wall, a floor and overlying fill. A dated object in the fill can constrain its formation without dating construction of the wall. Nearness to the floor does not show that it fell during occupation: establish whether it lay on the surface or arrived in later soil.

Several dates in a report do not necessarily compete to be “the right date”. They may concern different actions. Make a table of action, unit, material and method; only then ask whether results conflict. This separates a genuinely multiphase history from disagreement between measurements.

The sample must answer the question

A precise laboratory result cannot repair a weak archaeological association. For a sample in a mixed context, explain why it relates to the desired event. Selection, documentation and contamination are not merely technical issues: they determine what the number can mean. Further reading on sample and measurement considerations: M01.

Compare all measured samples with those included in the conclusion. If some were excluded, seek the reason. A justified exclusion may be appropriate; unexplained omission prevents assessment of dependence on selection. Readers can recognise that question without recalculating a model.

A broad range can be the correct result

Where evidence supports several possibilities, naming one year conceals the problem. A narrower range is not automatically better knowledge; examine how it was obtained. Rounding may aid timeline navigation but should not enter an object's description as though it were a precise measured result.

Return to the slingstone case: its research range includes contexts classified within two overlapping period frameworks. The evidence need not be altered to fit one period card. Retain the supported date and classification, then explain overlap. Precision includes declining a decision that the data do not require.