The Gold Standard · Archaeology
Make the decision path visible
Commutability is not a medal attached to a reference material forever. It is a demonstrated relationship: the material and authentic samples behave comparably across specified measurement procedures, within a stated criterion. Change the procedures, the software that turns signals into results or the matrices represented by the real samples, and that relationship may need to be tested again.
For archaeology, the issue matters whenever laboratories use reference materials to compare elemental, isotopic, molecular or other measurements across runs and institutions. A stable bottle does not guarantee a stable interpretation. The material may still be homogeneous and well characterised, yet cease to reveal the bias that matters for a new digestion, detector, calibration model or archaeological matrix.
Choose the route closest to your decision
Chapter contents
Chapters 1–3
Chapters 4–6
Chapters 7–9
Chapters 10–12
CHAPTER 1 OF 12 · DEFINE THE CLAIM
1. Write the commutability claim as a bounded statement
A useful claim names the reference material, the measurand, the participating measurement procedures, the authentic-sample population, the criterion and the period of evidence. The phrase commutable reference material without those boundaries invites overreach.
Separate commutability from stability and homogeneity. Stability asks whether the property changes during storage or use. Homogeneity asks whether sampled portions are sufficiently alike. Commutability asks whether relationships among procedures for the material resemble those for authentic samples. All three can matter, but passing one does not imply the others.
In an archaeological programme, authentic samples may include bone, enamel, sediment, ceramic, metal corrosion or plant remains. A conclusion demonstrated for one preparation and matrix family should not be silently extended to another.
CHAPTER 2 OF 12 · DEFINE THE CLAIM
2. Map every method and software version
Create a version map from sample preparation to reported result. Include digestion or extraction, instrument configuration, calibration function, blank correction, interference treatment, peak integration, isotope-ratio correction, database version and any post-processing script.
Software is part of the measurement procedure when it changes how signals become values. A new baseline algorithm can alter low-level results even though the instrument and reference material are unchanged. Record defaults as well as named versions, because an update can change behaviour without an operator editing the method file.
Classify changes as editorial, parameter-only or substantive. The classification is a hypothesis to test, not a permission to skip evidence.
Practical check
Write the decision, evidence and limit in three separate sentences. If one sentence depends on an unstated assumption, name it before continuing.
CHAPTER 3 OF 12 · DEFINE THE CLAIM
3. Define the authentic matrix population
Commutability evidence must represent the materials to which results will be applied. Choose authentic samples across the relevant concentration range, preservation states and matrix characteristics. Exclude samples only for stated technical reasons, not because they make the reference material look awkward.
A new excavation region, burial environment or conservation history can introduce matrices not represented in the original study. Soil mineralogy, diagenesis, binders and corrosion products can change extraction efficiency or interference. The question is not whether the new samples look unusual; it is whether the measurement relationship has changed.
Record how samples were selected. Convenience remnants may underrepresent difficult matrices and exaggerate generality.
| Element | Record |
|---|---|
| Measurand | Quantity and operational definition |
| Procedures | Method and software versions |
| Samples | Authentic matrix population |
| Criterion | Maximum permitted difference in bias |
CHAPTER 4 OF 12 · DETECT THE DRIFT
4. Set triggers for reassessment
Reassessment triggers include a new measurement procedure, a major software or firmware change, a replacement reagent system, a revised calibration model, a new reference-material lot and a materially different authentic-sample population.
Also monitor residuals and between-method differences over time. A slow change can appear before any formal method revision is recorded. Control materials help detect analytical drift, but only authentic samples can show whether the material-to-sample relationship remains comparable.
Do not wait for a failed proficiency round if the change is already visible in routine comparisons. A trigger starts an investigation; it does not automatically invalidate every prior result.
CHAPTER 5 OF 12 · DETECT THE DRIFT
5. Design an old-new bridge study
Measure the same authentic samples and reference-material portions using the old and new procedures where feasible. Balance run order, replicate enough to estimate repeatability and span the decision-relevant range. If the old system cannot be operated, preserve data and specimens that allow a partial bridge.
The study should compare relationships, not only mean levels. Plot procedure-to-procedure differences for authentic samples and locate the reference material against that distribution. A material can match the average while behaving differently across concentration or matrix strata.
Predefine the criterion and analysis before seeing the most favourable plot. Include uncertainty from value assignment, sample measurement and model fitting.
Practical check
Write the decision, evidence and limit in three separate sentences. If one sentence depends on an unstated assumption, name it before continuing.
CHAPTER 6 OF 12 · DETECT THE DRIFT
6. Work a matrix-drift example
A laboratory uses a bone-ash reference material to align phosphorus results from Method A and Method B. In the original study, the reference material's between-method bias differed from the authentic-bone bias by 0.3 percentage points, within a 1.0-point criterion.
After Method B receives a new spectral-correction model, authentic coastal bone samples show an average A-B difference of 1.8 points, while the reference material remains at 0.4. The resulting 1.4-point difference in bias exceeds the criterion. The material has not necessarily degraded; its commutability for that method pair and sample population is no longer supported.
Changed-context transfer check: inland cremated bone shows a 0.6-point difference and remains within criterion. The finding is therefore matrix- and procedure-specific. It would be wrong both to declare the material universally failed and to keep the old universal claim.
| Change | Question |
|---|---|
| Method update | Does sample behaviour change relative to the material? |
| Software update | Did signal processing alter results? |
| New matrix population | Is the original sample panel representative? |
| New lot | Does the replacement behave equivalently? |
CHAPTER 7 OF 12 · REPAIR COMPARABILITY
7. Analyse difference in bias and uncertainty
A common approach compares the bias between procedures for the reference material with the bias observed for authentic samples. The commutability criterion should reflect the intended use and the maximum additional bias that can be tolerated.
Model nonconstant differences when the relationship varies across the range. Inspect influence from individual samples and assess whether exclusions change the conclusion. Report an inconclusive region when uncertainty is too large to decide.
Statistical significance is not the same as fitness for purpose. A precisely estimated tiny difference may be acceptable, while a wide interval around a practically important difference may require more evidence.
CHAPTER 8 OF 12 · REPAIR COMPARABILITY
8. Choose among correction, restriction and replacement
If noncommutability is stable and method-specific, a validated correction with its own uncertainty may restore traceability for that procedure. The correction must be demonstrated on authentic samples and maintained as part of the calibration hierarchy.
If evidence supports only certain matrices or methods, narrow the certificate and laboratory instructions. Restriction is an honest scientific result, not a failure of ambition.
Replace the material or create a new matrix-matched material when correction is unstable, the authentic population has changed substantially, or the material cannot reproduce the relevant sample behaviour.
Practical check
Write the decision, evidence and limit in three separate sentences. If one sentence depends on an unstated assumption, name it before continuing.
CHAPTER 9 OF 12 · REPAIR COMPARABILITY
9. Protect historical comparability
Do not rewrite historical measurements merely because a current commutability conclusion changed. Identify the method and software version used for each result and the reference-material claim available at the time.
Where longitudinal comparison is important, remeasure retained specimens or archived extracts with both procedures if ethically and physically possible. Otherwise, use bridge uncertainty and segment the series. A conversion equation should not be applied beyond its validated matrices and range.
Archaeological material is finite. The value of consuming another aliquot must be weighed against the information gained and future analytical options.
| Finding | Action |
|---|---|
| Within criterion | Maintain bounded claim |
| Stable method-specific bias | Validate correction |
| Matrix-limited failure | Restrict claim |
| Unstable or broad failure | Replace or redesign material |
CHAPTER 10 OF 12 · GOVERN FUTURE USE
10. Monitor new matrix populations
Create a matrix-entry protocol for new projects. Compare routine authentic samples with the established commutability study and flag properties outside its coverage: mineral content, preservation, concentration range or preparation response.
Periodically include authentic-sample panels rather than relying only on the reference material. Rotate panels so that monitoring does not become blind to emerging matrices.
If sample access is limited, collaborate across laboratories using shared designs and transparent version metadata. Pooled evidence is only comparable when each procedure is sufficiently documented.
CHAPTER 11 OF 12 · GOVERN FUTURE USE
11. Control software lineage and issue clear reports
Archive raw signals, processed outputs, method files, software versions and transformation scripts. A final spreadsheet without lineage cannot reveal whether a change arose in the instrument, integration, correction or reporting layer.
Test software updates with locked datasets before deployment. Expected results should include difficult peaks, low concentrations and matrix interferences, not only clean reference runs. When proprietary software prevents full inspection, record configuration, vendor release notes and observed changes.
State where commutability was demonstrated, not demonstrated or remains inconclusive. Name the method pairs, authentic sample population, criterion, study date and software versions. Distinguish assigned-value uncertainty from commutability uncertainty, and keep superseded statements accessible with effective dates.
Practical check
Write the decision, evidence and limit in three separate sentences. If one sentence depends on an unstated assumption, name it before continuing.
CHAPTER 12 OF 12 · GOVERN FUTURE USE
12. Know when to stop extending the claim
Stop extending the original claim when new methods cannot be bridged, authentic matrices fall outside the study, correction varies unpredictably or uncertainty overwhelms the decision criterion.
The practical endpoint may be a family of matrix-specific materials, a new reference procedure or a reporting rule that prevents direct comparison. These are design decisions, not merely statistical adjustments.
The gold-standard habit is to preserve the boundary of the evidence. A stable material remains valuable even when its commutability claim must become narrower.
| Record | Purpose |
|---|---|
| Raw and processed data | Trace software effects |
| Authentic-sample panel | Represent intended use |
| Version map | Locate the change |
| Certificate history | Preserve interpretation over time |
Frequently asked questions
Can a stable reference material lose commutability?
Yes. Stability concerns the material over time; commutability also depends on the specified measurement procedures and authentic samples.
Does every software update require a full study?
Not automatically. Screen the update with locked data and bridging evidence, then scale reassessment to the observed and potential effect.
Can a correction solve noncommutability?
Sometimes, when the bias is method-specific, stable and validated on authentic samples. Its uncertainty must be included.
Should historical results be converted to the new method?
Only within a validated bridge. Otherwise preserve versioned segments and explain the comparability limit.
Continue through verified routes
Core routes
Archaeology subject root · Gold Standard of Education · Gold Standard directory · How X Works hub
Specialist routes
- Commutability, matrix mismatch and method-specific bias
- Reference-material lot changes
- Reference-material stability
- Aliquot homogeneity and depletion planning