A chromatic adaptation transform can move colour coordinates from one white-point condition toward another. An individual observer can still choose a different corresponding match because the model and the person are different sources of evidence.
This guide follows source and destination whites through response-space scaling, partial adaptation, inverse transformation, controlled matches and observer spread.
All numerical values are fictional teaching inputs. They are not a colour-critical production tolerance or a clinical colour-vision test.
The practical habit is to keep a small evidence receipt beside every answer. It names the object, reference condition, observation method, transformation and the nearest tempting claim that still does not follow.
Each changed-context check alters one consequential condition while the remaining basis stays fixed. This makes transfer visible: the reader can predict which field should change, which should remain stable and what new evidence would be required.
The article uses five routes so readers can enter by task, but the mechanism is cumulative. Definitions establish the objects, the worked case carries the arithmetic, evidence checks test rivals, transfer cases change the context and the final route repairs common shortcuts.
A useful comparison begins with a ledger rather than a conclusion. Put the two candidate quantities in separate columns, write the reference and observation rule for each, and leave an explicit blank when a field is unknown. A later correction can then change one field without silently changing the claim.
Keep three layers visible: what was directly observed, what was calculated or selected by a rule, and what action or interpretation is proposed. Evidence for the first two layers does not automatically justify the third. A careful report names the nearest rival explanation and one check that would distinguish it.
Before leaving the worked route, confirm that units and coordinates close, every derived value points back to a stored input and every linked article answers a materially different question. Read the conclusion alone. If it sounds universal, add the missing condition directly to the sentence.
This specialist guide belongs to the Visual Arts root. It also connects to The Gold Standard of Education | Learning That Transfers into Life and the How X Works hub. The worked cases are original teaching examples; their numbers are stipulated so that the reasoning can be checked.
eduKateSG · Visual Arts
Find your next learning step
Read the short answer first: a CAT is a named model with declared whites, observer basis and viewing conditions. Compare its output with individual matches as validation evidence; never relabel the computed coordinate as what every person sees.
ROUTE 1 · CHAPTERS 1–4
Define the quantities
Name the quantities, reference conditions and reporting rules.
ROUTE 3 · CHAPTERS 9–12
Audit the evidence
Check instruments, algorithms, assumptions and rival explanations.
ROUTE 5 · CHAPTERS 17–20
Report and continue
Repair shortcuts, answer questions and follow verified routes.
Chapter contents
Define the quantities · Chapters 1–4
1. Name model output and human match
2. Record source and destination whites
Work the mechanism · Chapters 5–8
5. Work the central cone-response case
6. Compute channel scaling explicitly
Audit the evidence · Chapters 9–12
9. Audit the standard observer
Transfer the reasoning · Chapters 13–16
13. Transfer to print and display whites
14. Transfer to incomplete adaptation
Report and continue · Chapters 17–20
17. Repair the white-balanced means accurate shortcut
18. Build a model-versus-observer ledger
A chromatic adaptation transform predicts corresponding colour coordinates across viewing illuminants under a defined model. An individual observer match is a person’s judgement under an experimental task. Agreement is evidence about model performance, not identity between computation and perception.
Write the evidence in the order a reader can audit: the Visual Arts quantity, its reference object, the observation or browser rule, the transformation, and the decision being made. Keeping those fields visible prevents a familiar label from hiding a changed basis.
A fictional grey patch under source white A is transformed to destination white D65. The CAT predicts XYZ 20.0, 21.0, 22.8; one observer adjusts a comparison to 20.4, 20.8, 23.3. Store both rows.
First-principles test: ask what physical, observational or algorithmic operation produces each value in this chapter. Then ask whether the second value could change while the first stays fixed. If it can, the two labels are not substitutes. For Visual Arts, this counterfactual is often clearer than a definition learned in isolation.
Keep the CAT inputs fixed and change the observer. The computed row stays fixed while the match can move.
Evidence check: another reader should be able to reproduce the narrow result from the stored definitions, inputs and intermediate steps, then identify the nearest stronger claim that the evidence does not support.
The numerical example is pedagogical and is not a colour-critical product specification.
A CAT needs the source white, destination white and adaptation assumptions. XYZ coordinates without their illuminants cannot support a corresponding-colour claim.
Trace the mechanism before interpreting the result. For Visual Arts, mark which value belongs to the source, which belongs to the measurement or layout algorithm, and which is a reported summary. A calculation is useful only when those roles remain attached.
Record source white Ws=(96.4,100.0,82.5) and destination Wd=(95.0,100.0,108.9) in the fictional case.
Practical transfer: build two records that share the headline number but differ in the condition discussed here. If the correct interpretation changes, the omitted condition is decision-relevant and belongs in the report. This test turns a vocabulary distinction into a reusable Visual Arts checking habit.
Swap the direction of adaptation. The scaling ratios invert; the original output cannot be reused.
Practical check: keep one counterexample beside the worked case. If the wording fails when that one condition changes, narrow the sentence until it describes only the evidence actually in hand.
Real measurements require calibrated spectral or tristimulus data and declared observer functions.
CHAPTER 3 OF 20 · Define the quantities
3. Separate tristimulus coordinates from appearance
XYZ values are standard-observer colourimetric quantities. A colour appearance model adds viewing-condition variables and perceptual correlates. A CAT is one component, not a universal description of how every person sees the patch.
Use a compact decision table with columns for condition, input, operation, result and limit. This makes the comparison reproducible and exposes where a missing field would force an assumption instead of an observation.
Place XYZ, CAT-adapted XYZ and observer match in three columns rather than calling them all “the perceived colour.”
Misconception check: a plausible output does not prove the right reference, denominator, ancestor, spectrum or observation rule was used. Reconstruct one intermediate step and one changed-context prediction. If either cannot be stated, keep the result provisional rather than strengthening its wording.
Hold XYZ fixed and darken the surround. Appearance can change without a new reflectance measurement.
Transfer check: repeat the reasoning with one consequential condition changed while holding the rest fixed. State what changes, what stays comparable and why the original label alone was insufficient.
CIE models have stated scopes and should be used with the appropriate publication.
Bradford, CAT02, CAT16 and other transforms use different matrices or response spaces. A report must name the transform and version rather than cite a generic “white balance.”
Preserve the raw inputs beside the derived result. If a later reader changes one condition, the original Visual Arts reasoning should still be recoverable without reverse-engineering a rounded headline or an unlabeled screenshot.
Use a stipulated three-channel teaching matrix and label it CAT-T. The calculation remains auditable without claiming CAT-T is a CIE recommendation.
Evidence design: retain both a positive case and a near-miss. The positive case shows when the rule applies; the near-miss changes one consequential condition and shows when the conclusion fails. Together they teach the boundary of the Visual Arts mechanism better than another unlabelled example.
Run the same whites through a second transform and preserve the different output.
Audit check: retain the source version, units, settings, exclusions and rounding rule. A future correction should be able to alter one field without rebuilding the entire explanation from memory.
Do not compare outputs from unnamed transforms as though rounding caused every difference.
The teaching transform maps the sample and the two whites into a response space, scales sample channels by destination-white/source-white ratios and maps back. The stages show where the white-point assumption enters.
Ask what would have to stay fixed for the comparison to be fair. Record those controls explicitly, then name the first uncontrolled condition that could reverse the conclusion. This is the practical transfer step, not an optional footnote.
Sample response is (18,22,24); source-white response is (95,102,80); destination-white response is (98,100,112). Scaled response is about (18.57,21.57,33.60).
Communication test: replace any sentence that says two things are “the same” or “different” with a sentence naming the operation, reference condition and supported consequence. The extra words are useful when they let a future reader repeat the decision without access to the original author.
Change only destination-white third channel from 112 to 96. Predict that the third scaled response falls while the first two stay fixed.
Decision check: end with an action or interpretation that follows from the stated conditions, followed by a boundary sentence explaining when a specialist rule, local procedure or additional measurement is needed.
The response triples are invented and should not be substituted for an official transform matrix.
Von Kries-style scaling multiplies each response channel by the ratio of destination to source white in that channel. The calculation is a model operation, not a measurement of an individual cone.
Write the evidence in the order a reader can audit: the Visual Arts quantity, its reference object, the observation or browser rule, the transformation, and the decision being made. Keeping those fields visible prevents a familiar label from hiding a changed basis.
Ratios are 98/95≈1.0316, 100/102≈0.9804 and 112/80=1.4. Apply them to the sample response before any inverse transform.
First-principles test: ask what physical, observational or algorithmic operation produces each value in this chapter. Then ask whether the second value could change while the first stays fixed. If it can, the two labels are not substitutes. For Visual Arts, this counterfactual is often clearer than a definition learned in isolation.
Set source and destination whites equal. All ratios become one and the transform should return the input apart from numerical error.
Evidence check: another reader should be able to reproduce the narrow result from the stored definitions, inputs and intermediate steps, then identify the nearest stronger claim that the evidence does not support.
Channel independence is a model assumption and does not settle every visual context.
After scaling, the inverse matrix returns adapted XYZ or another declared output. Rounding before inversion can create avoidable drift, so retain precision and the exact matrix.
Trace the mechanism before interpreting the result. For Visual Arts, mark which value belongs to the source, which belongs to the measurement or layout algorithm, and which is a reported summary. A calculation is useful only when those roles remain attached.
Store the scaled response to at least six digits, invert, then round the final XYZ for display.
Practical transfer: build two records that share the headline number but differ in the condition discussed here. If the correct interpretation changes, the omitted condition is decision-relevant and belongs in the report. This test turns a vocabulary distinction into a reusable Visual Arts checking habit.
Round each ratio to one decimal before inversion and quantify the extra error.
Practical check: keep one counterexample beside the worked case. If the wording fails when that one condition changes, narrow the sentence until it describes only the evidence actually in hand.
Numerical reproducibility requires matrix orientation, normalization and precision conventions.
A model can be assessed against corresponding-colour datasets or controlled matches. Residuals need direction, uncertainty and task conditions; one small difference does not prove universality.
Use a compact decision table with columns for condition, input, operation, result and limit. This makes the comparison reproducible and exposes where a missing field would force an assumption instead of an observation.
For five fictional observers, compute differences between CAT prediction and each matched XYZ. Report the distribution and individual rows, not only their mean.
Misconception check: a plausible output does not prove the right reference, denominator, ancestor, spectrum or observation rule was used. Reconstruct one intermediate step and one changed-context prediction. If either cannot be stated, keep the result provisional rather than strengthening its wording.
Add an observer whose match is farther away. The model output is unchanged while the evidence about population fit changes.
Transfer check: repeat the reasoning with one consequential condition changed while holding the rest fixed. State what changes, what stays comparable and why the original label alone was insufficient.
Statistical treatment must match sample size and experimental design.
The 2° and 10° standard colorimetric observers are defined functions for colourimetry, not literal average people who erase individual variability. The chosen observer affects the input XYZ.
Preserve the raw inputs beside the derived result. If a later reader changes one condition, the original Visual Arts reasoning should still be recoverable without reverse-engineering a rounded headline or an unlabeled screenshot.
Recalculate the patch from the same spectrum with stipulated 2° and 10° functions and keep two input rows before the CAT.
Evidence design: retain both a positive case and a near-miss. The positive case shows when the rule applies; the near-miss changes one consequential condition and shows when the conclusion fails. Together they teach the boundary of the Visual Arts mechanism better than another unlabelled example.
Keep the CAT unchanged but switch observer functions. A changed output begins upstream of the adaptation transform.
Audit check: retain the source version, units, settings, exclusions and rounding rule. A future correction should be able to alter one field without rebuilding the entire explanation from memory.
Use official CIE data and appropriate geometry in real work.
Some models include complete or partial adaptation. A degree parameter changes how far response ratios move from one toward the full white-point scaling.
Ask what would have to stay fixed for the comparison to be fair. Record those controls explicitly, then name the first uncontrolled condition that could reverse the conclusion. This is the practical transfer step, not an optional footnote.
Let D=0.8 in a stipulated interpolation. The third ratio becomes 1+0.8(1.4−1)=1.32 instead of 1.4.
Communication test: replace any sentence that says two things are “the same” or “different” with a sentence naming the operation, reference condition and supported consequence. The extra words are useful when they let a future reader repeat the decision without access to the original author.
Set D=0. The adapted response should remain the source response under the teaching rule.
Decision check: end with an action or interpretation that follows from the stated conditions, followed by a boundary sentence explaining when a specialist rule, local procedure or additional measurement is needed.
The correct adaptation formulation belongs to the selected model; do not mix formulas across models.
Adaptation depends on more than chromaticity. Adapting luminance, surround, field size and spatial structure can affect appearance and model parameters.
Write the evidence in the order a reader can audit: the Visual Arts quantity, its reference object, the observation or browser rule, the transformation, and the decision being made. Keeping those fields visible prevents a familiar label from hiding a changed basis.
Keep both whites fixed but change the fictional surround from average to dark. Record a new viewing-condition row before computing appearance correlates.
First-principles test: ask what physical, observational or algorithmic operation produces each value in this chapter. Then ask whether the second value could change while the first stays fixed. If it can, the two labels are not substitutes. For Visual Arts, this counterfactual is often clearer than a definition learned in isolation.
Change only field size and use a different standard-observer basis if the method requires it.
Evidence check: another reader should be able to reproduce the narrow result from the stored definitions, inputs and intermediate steps, then identify the nearest stronger claim that the evidence does not support.
A desktop mock-up cannot establish gallery or print-booth perception without controlled viewing.
CHAPTER 12 OF 20 · Audit the evidence
12. Audit measurement and display characterisation
Instrument calibration, flare, black level, spectral bandwidth and display profile can move measured coordinates. A model residual can therefore include measurement-system error.
Trace the mechanism before interpreting the result. For Visual Arts, mark which value belongs to the source, which belongs to the measurement or layout algorithm, and which is a reported summary. A calculation is useful only when those roles remain attached.
Measure a control tile before and after the session. A drift larger than the observer spread blocks a strong perceptual conclusion.
Practical transfer: build two records that share the headline number but differ in the condition discussed here. If the correct interpretation changes, the omitted condition is decision-relevant and belongs in the report. This test turns a vocabulary distinction into a reusable Visual Arts checking habit.
Keep observers fixed but correct the display profile. The apparent CAT error may shrink.
Practical check: keep one counterexample beside the worked case. If the wording fails when that one condition changes, narrow the sentence until it describes only the evidence actually in hand.
Traceability and uncertainty belong in colour-critical workflows.
A print viewed under one illuminant and a self-luminous display have different physical whites, surrounds and adaptation contexts. A shared D65 label does not make their appearance conditions identical.
Use a compact decision table with columns for condition, input, operation, result and limit. This makes the comparison reproducible and exposes where a missing field would force an assumption instead of an observation.
Transform the print measurement toward the display white, then compare under controlled soft-proof conditions.
Misconception check: a plausible output does not prove the right reference, denominator, ancestor, spectrum or observation rule was used. Reconstruct one intermediate step and one changed-context prediction. If either cannot be stated, keep the result provisional rather than strengthening its wording.
Change the booth illuminant while keeping the file values fixed. Recompute inputs and repeat the human comparison.
Transfer check: repeat the reasoning with one consequential condition changed while holding the rest fixed. State what changes, what stays comparable and why the original label alone was insufficient.
Soft proofing remains conditional on device characterisation and viewing control.
A viewer moving quickly between illuminants may not reach the model’s assumed adaptation state. The measured match can lag while a full-adaptation CAT output changes immediately.
Preserve the raw inputs beside the derived result. If a later reader changes one condition, the original Visual Arts reasoning should still be recoverable without reverse-engineering a rounded headline or an unlabeled screenshot.
Collect fictional matches after 5 s and 60 s. Keep the CAT output fixed and label time as an experimental variable.
Evidence design: retain both a positive case and a near-miss. The positive case shows when the rule applies; the near-miss changes one consequential condition and shows when the conclusion fails. Together they teach the boundary of the Visual Arts mechanism better than another unlabelled example.
Increase adaptation time. Predict convergence only as a hypothesis, not a guaranteed outcome for every observer.
Audit check: retain the source version, units, settings, exclusions and rounding rule. A future correction should be able to alter one field without rebuilding the entire explanation from memory.
Human experiments need ethical, repeatable protocols and adequate samples.
Two samples matching under one illuminant can diverge spectrally under another. A CAT that starts from tristimulus values cannot replace spectral evidence about metamerism.
Ask what would have to stay fixed for the comparison to be fair. Record those controls explicitly, then name the first uncontrolled condition that could reverse the conclusion. This is the practical transfer step, not an optional footnote.
Take two reflectance spectra with equal source XYZ, compute their destination XYZ spectrally, then apply the same appearance analysis.
Communication test: replace any sentence that says two things are “the same” or “different” with a sentence naming the operation, reference condition and supported consequence. The extra words are useful when they let a future reader repeat the decision without access to the original author.
Replace one spectrum with a physically identical duplicate. The metameric divergence disappears while observer variability can remain.
Decision check: end with an action or interpretation that follows from the stated conditions, followed by a boundary sentence explaining when a specialist rule, local procedure or additional measurement is needed.
Use the metamerism branch when spectral mismatch is the reader’s main question.
CHAPTER 16 OF 20 · Transfer the reasoning
16. Transfer to individual-observer variability
Lens, macular pigment, cone fundamentals, field size and task strategy can contribute to individual matches. A standard observer and a CAT provide common computation, not a guarantee that every person matches it.
Write the evidence in the order a reader can audit: the Visual Arts quantity, its reference object, the observation or browser rule, the transformation, and the decision being made. Keeping those fields visible prevents a familiar label from hiding a changed basis.
Plot each observer residual around the CAT prediction and retain repeat trials for within-person variability.
First-principles test: ask what physical, observational or algorithmic operation produces each value in this chapter. Then ask whether the second value could change while the first stays fixed. If it can, the two labels are not substitutes. For Visual Arts, this counterfactual is often clearer than a definition learned in isolation.
Average the observers and hide the spread. Explain which decision-relevant information is lost.
Evidence check: another reader should be able to reproduce the narrow result from the stored definitions, inputs and intermediate steps, then identify the nearest stronger claim that the evidence does not support.
Clinical interpretation of colour vision lies outside this article.
CHAPTER 17 OF 20 · Report and continue
17. Repair the white-balanced means accurate shortcut
A neutral-looking white after a CAT does not prove all colours, observers and contexts are accurately predicted. The transform controls a defined adaptation relationship.
Trace the mechanism before interpreting the result. For Visual Arts, mark which value belongs to the source, which belongs to the measurement or layout algorithm, and which is a reported summary. A calculation is useful only when those roles remain attached.
Rewrite “the image is perceptually corrected” as “the image was transformed from named source white to named destination white using CAT16 assumptions; observer validation is separate.”
Practical transfer: build two records that share the headline number but differ in the condition discussed here. If the correct interpretation changes, the omitted condition is decision-relevant and belongs in the report. This test turns a vocabulary distinction into a reusable Visual Arts checking habit.
Test a saturated patch and a neutral patch. Similar white agreement can coexist with different residuals.
Practical check: keep one counterexample beside the worked case. If the wording fails when that one condition changes, narrow the sentence until it describes only the evidence actually in hand.
Accuracy claims need task-specific validation and tolerances.
The ledger stores spectra or XYZ inputs, observer function, whites, transform, viewing conditions, CAT output, individual matches, repeats and residuals. It prevents a mean human result from overwriting individual evidence.
Use a compact decision table with columns for condition, input, operation, result and limit. This makes the comparison reproducible and exposes where a missing field would force an assumption instead of an observation.
Ask a second analyst to reproduce the transform and a second experimenter to reproduce the match task from the same receipt.
Misconception check: a plausible output does not prove the right reference, denominator, ancestor, spectrum or observation rule was used. Reconstruct one intermediate step and one changed-context prediction. If either cannot be stated, keep the result provisional rather than strengthening its wording.
Remove the transform version. The numerical result becomes difficult to audit even if screenshots remain.
Transfer check: repeat the reasoning with one consequential condition changed while holding the rest fixed. State what changes, what stays comparable and why the original label alone was insufficient.
Data protection and experimental governance apply when collecting participant data.
A defensible conclusion says what the transform predicted, under which whites and viewing assumptions, how observers were tested and how variable their matches were. It avoids treating a model output as a biological measurement.
Preserve the raw inputs beside the derived result. If a later reader changes one condition, the original Visual Arts reasoning should still be recoverable without reverse-engineering a rounded headline or an unlabeled screenshot.
Write: “CAT-T predicted the stated destination XYZ; five observers produced a spread around it under the controlled task, so the model is a common reference rather than a universal individual match.”
Evidence design: retain both a positive case and a near-miss. The positive case shows when the rule applies; the near-miss changes one consequential condition and shows when the conclusion fails. Together they teach the boundary of the Visual Arts mechanism better than another unlabelled example.
Change one observer function or surround and identify which parts of the conclusion must be recomputed.
Audit check: retain the source version, units, settings, exclusions and rounding rule. A future correction should be able to alter one field without rebuilding the entire explanation from memory.
The fictional data support learning only.
Standard observers versus individual matches, metamerism versus colour constancy and perceptual colour difference versus luminance contrast answer different questions around this model.
Ask what would have to stay fixed for the comparison to be fair. Record those controls explicitly, then name the first uncontrolled condition that could reverse the conclusion. This is the practical transfer step, not an optional footnote.
Follow the standard-observer branch for population functions, the metamerism branch for spectral pairs and this branch for adaptation-model output.
Communication test: replace any sentence that says two things are “the same” or “different” with a sentence naming the operation, reference condition and supported consequence. The extra words are useful when they let a future reader repeat the decision without access to the original author.
Change the task to simultaneous contrast without changing illumination and use the dedicated route.
Decision check: end with an action or interpretation that follows from the stated conditions, followed by a boundary sentence explaining when a specialist rule, local procedure or additional measurement is needed.
Grouped reading should preserve each article’s measurement and claim boundary.
Continue through the verified reading route
Standard-observer values and individual matches
Standard-observer values and individual matches. Use this for the construction and limits of standard colorimetric observers.
Metamerism and colour constancy
Metamerism and colour constancy. Use this when spectral matches and stable object appearance are being combined.
Perceptual colour difference and luminance contrast
Perceptual colour difference and luminance contrast. Use this when colour-difference and light-dark contrast metrics answer different questions.
Frequently asked questions
Is a CAT output a measurement of one person?
No. It is a model output on a declared colorimetric and viewing basis.
Why can observers disagree with one another?
Individual physiology, task, field, adaptation and repeatability can all contribute; preserve individual trials.
Does a correct white guarantee every colour is correct?
No. White-point adaptation is one part of a colour-appearance workflow and needs task-specific validation.
Authoritative sources and further reading
CIE 160:2004. The official CIE review describes experimental datasets and multiple chromatic adaptation transforms.
CIE 015:2018 Colorimetry. The official publication covers standard observers, illuminants, tristimulus values and advanced colorimetry.
ISO/CIE 11664-1:2019. The official standard defines CIE standard colorimetric observers.
CIECAM16. The official CIE publication describes a colour appearance model containing a chromatic adaptation transform.
Return to the Visual Arts subject guide, The Gold Standard of Education, or the How X Works hub.