Scientific testing rarely answers the complete collector question, “Is this object genuine?” It normally answers something narrower: what material is present, whether a substance is detectable, when a sampled material probably formed or was last heated, how an object is constructed, whether two samples are distinguishable, or whether a claimed date is contradicted.
The collector's task is therefore not merely to obtain a test. It is to connect the observation to the right material, component, historical fact and market claim without skipping the assumptions in between. The safest final opinion is often layered rather than binary: original core, later repair, unresolved signature, modern mount, plausible attribution and overstated sale description.
Central authentication principle
A test result authenticates only what it actually measures.
Every broader conclusion depends on sample relevance, material identification, historical interpretation, alternative explanations and evidence from the rest of the object.
Orientation
A result is not the same as a conclusion
Scientific language becomes unreliable when observations, identifications and authentication claims are collapsed into one statement. Read the chain of inference one step at a time.
1
Instrument observation
The spectrum, image, count, ratio or measurement recorded by the instrument. Example: peaks associated with titanium are present.
2
Material interpretation
A scientific identification inferred from the observation. Example: the white material probably contains titanium dioxide.
3
Historical interpretation
A conclusion about availability, manufacture or chronology. Example: this formulation may be inconsistent with manufacture before a particular period.
4
Object-level interpretation
A conclusion about the tested component or layer. Example: the measured area probably contains later material.
5
Authentication conclusion
A conclusion about the complete market claim. Example: the object is a modern forgery. This final step normally requires evidence beyond the instrument result alone.
The dangerous leap
A modern substance may indicate forgery, but it may also be restoration, retouching, contamination, a replacement part, a later inscription, a mounting material, varnish, adhesive or a genuine old object assembled with later additions.
The narrowest defensible conclusion should be stated first. Broader conclusions should be earned by additional evidence, not implied by the authority of the instrument.
Evidential meaning
What a scientific result can establish
The same laboratory phrase can carry very different authentication weight. Distinguish exclusion, compatibility, association and dating before deciding what the result changes.
Strong exclusion
What the evidence means
A securely identified feature is difficult or impossible to reconcile with the claim, such as a post-dating polymer in an original structural layer.
Collector risk
The result may still be misused if the anomalous material actually belongs to repair, coating, mounting or contamination.
Proportionate action
State precisely which claim, component and date are contradicted before extending the conclusion to the whole object.
Compatibility
What the evidence means
No contradiction was found within the tested scope. The material, process or date could fit the proposed history.
Collector risk
Compatible period materials can also be used in later manufacture, restoration or deliberate deception.
Proportionate action
Record the result as supportive but non-exclusive. Ask what alternative explanations would produce the same finding.
Positive association
What the evidence means
The result links the sample to a source, workshop, process, batch, quarry, mill, production period or comparison group.
Collector risk
A laboratory ‘match’ may only mean the samples were not distinguishable by the chosen method.
Proportionate action
Ask how rare the profile is, how large the comparison population was and whether unrelated objects commonly produce the same result.
Dating evidence
What the evidence means
The test estimates a material event such as growth, death, firing, heating, exposure, deposition or cooling.
Collector risk
The dated event may predate assembly by decades or centuries, especially where old material or ancient fragments were reused.
Proportionate action
Connect the dated event to manufacture explicitly and preserve the complete probability range, not a convenient single year.
Question design
Begin with the proposition the test was meant to answer
A technically excellent test can still be irrelevant if the question was poorly framed. Separate source, activity and whole-object authentication propositions.
Source proposition
Who or what produced the tested material?
Example: did this ink come from the same formulation or source as the reference ink?
Activity proposition
What happened to the object, component or layer?
Example: was the signature added after the document had aged?
Authentication proposition
What does the whole body of evidence say about the market claim?
Example: is this a genuinely signed first edition rather than a later-added signature?
Better questions produce better evidence
Is the ink chemically compatible with the claimed date?
Is it distinguishable from ink on a securely genuine comparison document?
Does stroke sequence show whether the signature preceded or followed the printed text?
Is there evidence of tracing, transfer, retouching or surface disturbance?
Are the signature and document likely to have been created in the same period?
Sampling
Understand exactly what was tested
A result applies first to the measurement location or sample. It should not be borrowed by untested layers, components or the complete object without justification.
Questions about location
Was the surface, bulk material or a buried layer examined?
Was one location tested or were several areas mapped?
Could the reading include coating, corrosion, dirt, restoration or mounting material?
Was the beam or sampling area larger than the feature under investigation?
Was the object tested through glass, varnish, plastic or packaging?
Questions about representativeness
Is the item homogeneous or made from multiple chronological units?
Does the sample belong to the disputed feature?
Could it come from a known repair or detachable component?
How large was the sample and why was that point selected?
Could a genuine old fragment have been inserted into a later assembly?
Collector scenario: one bronze, several histories
A bronze may contain an old body, a nineteenth-century repair, a modern patina, recent solder, soil deposits and an old mounting pin. A single reading can be accurate for the measured spot yet misleading as a description of the complete sculpture.
Composite objects should be interpreted as maps of components, joints, coatings and suspicious interventions. The correct conclusion may be “genuine core with later additions” rather than a forced choice between wholly real and wholly fake.
Reading the report
Separate raw data, processing, scientific interpretation and authentication opinion
A useful report allows the reader to see how the analyst moved from signal to conclusion and where judgement entered the process.
Raw or near-raw evidence
Spectra, chromatograms, elemental counts, isotope ratios, radiographs, fluorescence images, micrographs, diffraction patterns, signal intensities and calculated dates.
Statements such as probable lead white, polyester fibre detected, spectrum resembles shellac, or alloy compatible with historical brass.
Authentication interpretation
Statements about whether a tested feature is later, inconsistent, compatible, non-diagnostic or unable to resolve the market claim.
Confidence and limits
Identification language is graded, not binary
Terms such as identified, matched, detected and consistent with are often used loosely. Their evidential strength depends on signal quality, reference data, alternatives and sampling.
Confirmed identification
A strong diagnostic signal, appropriate reference data, adequate signal-to-noise and plausible alternatives excluded, sometimes with confirmation by a second method.
Probable identification
The best explanation of the data, but one or more material, signal, reference or sampling uncertainties remain.
Class-level identification
The method identifies a broad family, such as a protein, drying oil, synthetic resin or iron-containing pigment, but not a precise source or recipe.
Trace detection
A small signal that may reflect a genuine minor component, contamination, handling, carry-over, restoration, environmental deposition or processing artefact.
Not detected
The substance was not observed under the conditions used. It may still be present below the detection limit, beneath another layer or outside the sampled location.
Inconclusive
The evidence cannot distinguish the competing explanations reliably. This may be the most scientifically responsible result.
“Not detected” is not “absent”
A substance may be below the detection limit, beneath another layer, outside the sampled location, masked by overlap, degraded, blocked by a coating or poorly suited to the method.
The responsible wording is: not observed under the stated conditions.
Trace presence is not automatically decisive
A marginal signal may be a genuine ingredient, but it may also be contamination, carry-over, handling residue, environmental deposition, restoration material or processing artefact.
Replicate the finding and explain its location before using it as a chronological boundary.
Measurement
Uncertainty, detection limits and numerical theatre
Decimal places can create an appearance of certainty that the method, sample and interpretation do not deserve.
A more honest numerical result
“8.1% tin” looks precise. A more informative statement might be: “Estimated tin concentration 8.1%, expanded uncertainty ±1.2%, subject to surface corrosion and calibration limitations.”
If the authentication argument turns on whether the alloy is above or below 8%, the uncertainty may prevent a reliable classification.
Precision
Repeated measurements are close to one another.
Accuracy
The result is close to the relevant true value.
Reproducibility
Different operators or laboratories obtain comparable results.
When a weak signal should not carry the whole case
It lies close to the method's detection or reporting threshold.
It appears at only one location and is not replicated.
The peak, feature or match has plausible alternatives.
Contamination and restoration have not been considered.
The conclusion depends on a threshold narrower than the stated uncertainty.
Comparison evidence
A match is only as strong as the reference population
Two samples can be indistinguishable by one method without coming from the same source, workshop, batch or object.
Reference questions
Is the reference unquestionably authentic and securely provenanced?
Has it been restored, contaminated or altered?
Is the reference population large and representative?
Were the same method, settings and processing used?
Does the database include known reproductions and fakes?
What “match” might mean
Within the same broad material class
Compositionally similar
Statistically consistent
Indistinguishable by the selected method
Possibly from a common source or production batch
Circular authentication
Poorly provenanced objects may be accepted as genuine, entered into a reference database and then used to authenticate later items because they match. The comparison appears scientific while the underlying assumption has never been independently secured.
Ask the laboratory to complete this sentence: “The samples match with respect to ______, using ______, within ______ uncertainty, and the frequency of such a match among unrelated materials is ______.”
Inference
Compare competing explanations, not just favourable and unfavourable results
Scientific evidence gains meaning by asking how probable the finding would be under each plausible history of the object.
Example: modern pigment found
Hypothesis A
The object is a modern imitation.
Hypothesis B
The object is old but was restored with modern pigment.
The pigment result may support both. Location, stratigraphy, extent, crack relationships, conservation records and imaging may be needed to distinguish them.
Weak separation
Old paper is expected in a genuine eighteenth-century letter, but also plausible in a forgery written on antique blank paper. The result changes the assessment only slightly.
Stronger separation
A post-dating colorant embedded within the original ink stroke is much more expected under a later-addition hypothesis than under complete contemporaneity.
Method-specific interpretation
What major test types can and cannot tell you
The method defines the kind of evidence available. Each technique also has characteristic blind spots, interferences and sampling problems.
X-ray fluorescence (XRF)
What it may show
Elemental composition in metals, pigments, coatings and corrosion products
Differences between measured areas
Possible alloy, filler or pigment-related elements
Interpretive cautions
Often surface-sensitive and vulnerable to corrosion, plating, patina and coatings
Detecting an element does not automatically identify the complete compound
Geometry, calibration, overlapping peaks and limited sensitivity to light elements matter
Species or source identification does not prove maker attribution
Trade, recycling and imported materials may weaken geographic inferences
Chronology and intervention
Modern material, old material and the history between them
A material date becomes authentication evidence only when its location, function, availability and relationship to manufacture are understood.
Interpreting a modern-material finding
Confirm the identification and exclude plausible alternatives.
Locate it: original structure, surface, repair, coating, mount or contamination.
Establish whether its historical availability is genuinely diagnostic.
Determine whether it is original to the questioned feature.
Assess whether it is isolated or widespread.
Check documented treatment and restoration history.
State the exact claim contradicted: date, originality, signature, untouched condition or whole-object authenticity.
Interpreting an “old enough” result
Identify the dated event: growth, death, firing, exposure, deposition or manufacture.
Ask how closely that event precedes object creation.
Consider reuse of old timber, paper, canvas, metal, parchment or ceramic fragments.
Consider a modern assembly around a genuine old component.
Check whether the date range is narrow enough to distinguish the competing claims.
State what later manufacture remains possible.
Earliest-possible-date principle
A reliably identified later material in an original feature may provide a terminus post quem: the feature cannot reasonably predate the material's availability. This requires confidence in identification, historical availability, sample relevance and exclusion of restoration or contamination.
Latest-possible-date problem
Period-compatible material rarely proves that manufacture did not occur later. Old stock, recreated recipes, antique blanks and recycled components may all survive into modern use.
Converging evidence
Multiple methods help only when they add genuinely independent information
Confidence grows when different methods address different parts of the claim and converge with provenance, construction and specialist examination.
Visual and microscopic inspection establish where anomalies occur.
Imaging reveals hidden structure and guides targeted sampling.
Elemental and molecular analysis identify different aspects of materials.
Cross-sections place materials in chronological sequence.
Manufacturing history tests whether the result is historically significant.
Provenance and securely attributed comparisons provide independent context.
Correlated evidence
Five tests may add little if they all measure the same element, rely on the same questionable sample, share the same database or are affected by the same contamination.
Independent convergence
Two methods may be much stronger if one establishes material identity and another independently establishes layer sequence, while provenance and construction point in the same direction.
Disagreement and quality control
What to do when results conflict
Contradictory findings are not resolved by choosing the result that supports the preferred attribution. First determine whether the tests are genuinely addressing the same material and question.
Confirm whether the tests measured the same component, layer and property.
Compare sampling locations and surface-versus-bulk sensitivity.
Identify which method is more specific and which result is better replicated.
Review uncertainty, detection limits, calibration and known interferences.
Ask whether the object has multiple historical phases that allow both results to be true.
Seek independent review where the consequence is substantial.
Replication
Repeat measurements, sample more than one location and use a complementary method.
Blind interpretation
Where feasible, code samples or separate initial analytical review from market expectations.
Independent laboratory
Use another competent laboratory for high-value, litigated or reputation-sensitive disputes.
Specialist threshold
When laboratory competence and chain of custody become decisive
A sophisticated instrument does not guarantee a useful authentication opinion. Heritage materials, layered objects and disputed property require specialist competence and disciplined documentation.
Competence to look for
Experience with aged, degraded and conservation-treated materials
Understanding of layered and composite collectibles
Validated procedures, calibration and suitable reference standards
Contamination control and transparent reporting
Knowledge of historical manufacture and method limitations
Chain-of-custody essentials
Object identity, condition and photographs on receipt
Who selected and removed the sample
Exact location, size and authorisation for destructive work
Packaging, seals, storage and transfers between parties
Disposition of unused sample and ownership of data
Escalate before testing when
Sampling may affect condition, appearance, legal value or marketability.
The item is highly valuable, disputed, insured, litigated or culturally sensitive.
The proposed method is destructive or difficult to repeat.
The object is composite and sample representativeness is uncertain.
The laboratory is technically capable but lacks relevant heritage-material experience.
Probability and language
Read numbers and conclusion vocabulary at the right level
A probability range for a date is not a percentage chance that the object is authentic, and statistical significance is not automatically authentication significance.
Language that should raise concern
“Scientifically proven authentic”
“100% genuine” or “the machine cannot be wrong”
“Exact chemical match” without defined criteria
“The age test passed” without stating what was dated
“No modern materials found” without methods and detection limits
“Dated to 1642” where the method produced a broad probability range
More defensible vocabulary
Strongly supports or strongly contradicts
Supports or materially increases plausibility
Consistent with or compatible with
Cannot distinguish between competing explanations
Non-diagnostic, insufficient data or limited by sampling
Excludes the proposed date for the tested component
“Inconclusive” is not a failed result. It may be the correct conclusion when the evidence does not separate the plausible explanations reliably.
Collector decision model
Translate each finding through four disciplined questions
This framework prevents raw observations from becoming inflated market claims.
1. What was observed?
The direct signal, image, count, structure, date range or measurement.
2. What does it probably identify?
The material, process, source class or chronological event inferred from the data.
3. What historical fact follows?
Availability, manufacture, source, sequence or date implications, including uncertainty.
4. Which authentication claim is affected?
The exact statement supported, weakened, contradicted or left unresolved.
Assign the finding a strength—strong, moderate, limited, neutral, contradictory or unresolved—then integrate it with provenance, construction, comparison, documentary evidence, condition and other genuinely independent tests.
Applied judgement
Example interpretations across collectible categories
The examples below show how the same discipline works across paper, books, ceramics, metal, paintings and documents.
A nineteenth-century poster
Finding
Microscopy reveals modern inkjet droplets in the main image.
Responsible interpretation
This strongly contradicts the claim that the printed image is nineteenth century.
It does not determine whether the paper itself is old.
An old-paper modern reproduction remains possible.
A signed first edition
Finding
The paper, binding and signature ink are broadly compatible with the period.
Responsible interpretation
The book materials are not contradicted by the test.
The result does not prove that the named person signed it.
A later signature using historically compatible ink remains possible.
An ancient ceramic figure
Finding
Thermoluminescence supports ancient firing for a sample taken from the torso.
Responsible interpretation
The sampled ceramic is likely old.
The head, arms, surface decoration and joins require separate examination.
A modern composite containing an ancient torso remains possible.
A bronze sculpture
Finding
XRF shows an alloy compatible with historical bronze.
Responsible interpretation
No elemental contradiction was identified at the measured locations.
Similar alloys can be reproduced today.
Casting structure, patina, toolmarks, repairs and provenance remain necessary.
A restored painting
Finding
A modern pigment appears only in the uppermost retouching layer.
Responsible interpretation
This is evidence of later restoration.
It is not, by itself, evidence that the underlying painting is modern.
The market description should distinguish original work from later intervention.
A historical document
Finding
One ink entry contains a dye introduced after the document's claimed date.
Responsible interpretation
The entry is strongly indicated to be later.
The underlying document may still be genuine.
Its importance depends on whether the later entry creates the attribution or value.
Documentation
Minimum contents of a useful scientific report
A one-line certificate cannot preserve the reasoning, limits or connection between the test and the exact object.
Object and scope
Full description, dimensions, identifiers and photographs
Submitting party, condition and component relationships
Precise questions and claims investigated
Components included, excluded and left untested
Examination and methods
Sample locations, rationale, size and surface preparation
Instrument, method, settings, calibration and reference standards
Software, libraries, qualitative or quantitative status
Detection limits, uncertainty and relevant limitations
Results and interpretation
Representative raw or processed data
Replicate measurements and confidence of identification
Historical significance and alternative explanations
Restoration, contamination and sample-relevance considerations
Conclusion and authorship
Narrow, scoped conclusions and evidential strength
Untested questions and recommended follow-up
Analyst, qualifications, laboratory, date and signature
Conflicts, commercial relationships and reliance limitations
A useful negative conclusion is: “No material inconsistency was identified in the examined locations using the methods described.” It is not equivalent to “the object is authentic.”
Before commissioning work
Questions a collector should ask the laboratory
What exact question can the test answer?
What can it not answer?
What part of the object will be tested?
Why is that location representative?
Is the method surface-sensitive or bulk-sensitive?
Is the result qualitative or quantitative?
What detection limits and uncertainty apply?
Which references or databases will be used?
Can restoration or contamination produce the same result?
Will several locations or replicate measurements be used?
Can another method confirm the finding?
Will spectra, images or numerical data be supplied?
How will alternative explanations be addressed?
Does the analyst authenticate objects or only report analytical findings?
Does the laboratory know this material and collectible category?
Will destructive sampling affect condition or value?
Who owns the sample and underlying data?
Can a future buyer rely on the report?
Reviewing an existing report
Check that the report still belongs to the object being offered
A genuine laboratory report can be misapplied to a different object, an altered item or a component that has since been replaced.
Does the photograph unmistakably match the offered object?
Are serials, dimensions and identifying marks consistent?
Are sample locations shown?
Was the tested component later replaced or altered?
Does the conclusion exceed the method's scope?
Was “consistent with” converted into certainty?
Are negative findings treated as proof of absence?
Was only one convenient probability range quoted?
Were known restorations considered?
Are the analyst and laboratory identified?
Are underlying results available?
Is the report transferable or addressed only to the original client?
Common failures
Interpretive habits that repeatedly create false confidence
Machine infallibility
Treating instrumental output as objective truth without considering sampling, processing and interpretation.
Single-test absolutism
Believing one method can settle a complex attribution or a composite object's entire history.
Surface-to-whole extrapolation
Applying a surface-sensitive reading to the bulk object.
Component-to-object extrapolation
Dating one old fragment and authenticating the whole assembly.
Compatibility inflation
Turning ‘could be period’ into ‘is genuine’.
Anomaly inflation
Turning one later feature into proof that every component is fake.
Restoration blindness
Ignoring legitimate conservation, repair and historic intervention.
Contamination blindness
Treating trace surface material as intentional original composition.
Reference overconfidence
Assuming a library hit or comparison match is unique.
Date-point illusion
Presenting a probability distribution as one exact year.
Selective reporting
Publishing favourable findings while omitting anomalies and limitations.
Decimal-place theatre
Using numerical precision to imply certainty the evidence does not support.
Final opinion
The best conclusion is often layered
A responsible authentication opinion describes the condition of the evidence rather than forcing every part of the object into one label.
Object body: probably period.
Surface: substantially restored.
Signature: unresolved.
Mount: modern.
Provenance: incomplete.
Attribution: plausible but unproved.
Market description: overstated.
This is more useful to a collector, insurer or future buyer than a bare certificate reading “real” or “fake”, because it preserves what is supported, what is altered and what remains uncertain.
Core collector rules
A test result authenticates only what it actually measures.
A sample does not automatically represent the whole object.
Material identification and historical interpretation are separate steps.
“Consistent with” is compatibility, not proof.
“Not detected” is not the same as absent.
An old component does not prove an old assembly.
A modern component does not always prove a modern object.
Uncertainty, detection limits and alternative explanations matter.
Reference samples must themselves be trustworthy.
Independent convergence is stronger than repeated versions of the same test.
Scientific evidence should be integrated with provenance, construction and specialist examination.
A responsible conclusion states exactly what remains unknown.
Final authentication question
How much more or less plausible does this specific result make each competing explanation of the object, given what was tested, how it was tested, the uncertainty involved and all the other available evidence?
