Short answer
Professional codes separate preventing change, stopping active damage, and restoring appearance, and they hold the last to the strictest tests: an addition should be documented, detectable on examination, removable, and must not falsify what the object is. Reversibility is an aim for what conservators add, not a property of the damage. In Colourwise's degradation dataset none of the 51 recorded colour changes is classed as fully reversible and 37 are classed as irreversible, which is why so much recent work restores colour in light or on a screen instead of on the object.
The terminology adopted by ICOM's Committee for Conservation in 2008 divides conservation into three. Preventive conservation covers measures that avoid or minimise future deterioration; they are indirect, act on the surroundings rather than the object, and do not modify its appearance — light budgets, humidity control and pollution filtering belong here. Remedial conservation acts directly on an object to arrest damage or reinforce it. Restoration acts directly on a single, stable item to help it be appreciated, understood and used; retouching a painting is the resolution's own example, and it notes that varnish removal can count as both restoration and remedial conservation. The three are distinguished by aim, and all are required to respect the significance and physical properties of the item. The division matters for colour because almost every colour change on this site is addressed, in practice, by the first category: slowing what cannot be undone.
Where colour is added to fill a loss — retouching, inpainting, toning a repair — the American Institute for Conservation's guideline on compensation for loss sets four conditions. The intervention should be documented in the treatment records. It should be detectable by common examination methods, so that a later examiner can tell new from old. It should be reversible. And it should not falsely modify the known aesthetic, conceptual and physical characteristics of the object, especially by removing or obscuring original material. A neighbouring guideline asks that materials and methods be weighed against their potential adverse effects on future examination, scientific investigation and treatment. Retouching confined to losses, in a medium that can be taken off without disturbing the original, is how those tests are usually met; repainting faded original paint to brighten it would fail them.
For buildings and monuments the Venice Charter of 1964 put the limit in a phrase: restoration 'must stop at the point where conjecture begins'. It is to be based on respect for original material and authentic documents, and replacements of missing parts must integrate harmoniously with the whole yet be distinguishable from the original, 'so that restoration does not falsify the artistic or historic evidence'. Applied to colour, that is the difficulty with repainting a façade or a statue in a reconstructed scheme. Pigment traces on the Parthenon Sculptures show that Egyptian blue was used and where; they do not supply the full scheme, and a painted reconstruction has to commit to a colour on every surface. Museums have answered by painting replicas rather than originals.
The word describes what a conservator adds, and even there it is an ideal. Cleaning cannot be reversed. A polymer formed inside a paint layer to hold it together — the approach reported for the Terracotta Army's polychromy — cannot be extracted again. The chemical conversion of blackened lead white has been reassessed over a century precisely because of what it does to the paper underneath. The damage itself is rarely reversible at all: of the 51 changes in the dataset, 37 are recorded as irreversible, 10 as partly reversible and 4 as disputed, and the partly reversible ones are mostly cases such as Prussian blue, whose colour partly returns by itself in air, or coatings that can be removed at some risk. The table lists every record that is not simply irreversible, with the reason.
The clearest recent response to these constraints leaves the object alone. For Rothko's Harvard Murals, faded beyond exhibition by daylight, Harvard Art Museums projected a calculated compensation image onto each canvas so that it appeared in its original colours; the authors called it inpainting with light and set it explicitly beside conventional cleaning and inpainting. The projectors were switched off for a period each day so that visitors could see the paintings as they are. At Hampton Court a tapestry was shown in its estimated original colours the same way. Such a treatment is documented, detectable — it vanishes at a switch — and wholly reversible, and it removes nothing. What it cannot escape is the charter's other test: the target colours are an estimate, and the museum presented the project as something to be studied and debated.
| Document | Issued by | Applies to | What it asks, in summary |
|---|---|---|---|
| Terminology to characterize the conservation of tangible cultural heritage (2008) | ICOM-CC | All tangible heritage | Separates preventive conservation, remedial conservation and restoration by aim; all must respect the item's significance and physical properties. |
| Code of Ethics and Guidelines for Practice, guideline 23 | American Institute for Conservation | Members' treatment of cultural property | Compensation for loss is documented, detectable, reversible, and does not falsely modify the object or obscure original material. |
| Venice Charter, Articles 9 and 12 (1964) | ICOMOS | Monuments and sites | Restoration stops where conjecture begins; replacements are harmonious but distinguishable from the original. |
| Colourant and change | Verdict | Why | What is documented as done |
|---|---|---|---|
| Zinc yellow (zinc potassium chromate) in oil: Yellow, green-yellow and orange strokes have become ochre, olive green and reddish brown. | Partly reversible | In the laboratory, dichromate in a paint film was turned back to yellow chromate by alkaline gases. That is a demonstration of the chemistry, not a treatment; the Cr(III) step is not reported as reversible. | No treatment of the painting is reported. Its earlier appearance was simulated digitally from spectral measurements. |
| Prussian blue, especially extended or mixed with white: Pale blues fade towards grey or greenish grey. | Partly reversible | Colour returned in the dark only after short exposures; after long exposure it did not, and further light caused more irreversible loss. | None reported; the study characterises which paints are at risk. |
| Dammar and mastic picture varnish: A clear coating turns yellow, then brown, veiling every colour beneath. | Partly reversible | The varnish itself does not recover. The layer is designed to be removable, but removal is conservator's work with a risk to the paint. | Conservators remove or reduce discoloured varnish, traditionally monitoring progress under UV; CCI stresses that removal is not always advisable or necessary. |
| Modern cadmium yellow and French ultramarine oil paints with magnesium carbonate: Crystals on the surface of the paint. | Partly reversible | No treatment is reported. The salts are water-soluble, but the study describes such paintings as water-sensitive. | None reported; the finding points to humidity control. |
| Cinnabar (mercury sulfide) on lime mortar: Deep red turns grey or black. | Disputed | The black coating was found lying over intact cinnabar, while the chlorine route alters the pigment itself. The study reports no treatment, so recoverability is not established. | None reported; the study separates two causes that look alike. |
| Lead white on plaster: White passages turn brownish and dark. | Partly reversible | The authors report a procedure that reconverted the brown product to white. It is a specialist chemical treatment and is not described here. | Chemical reconversion by conservators is reported in the literature; a later study of a different pigment shows such treatments can leave unstable products. |
| Red lead (minium) on plaster: Orange-red turns brown-black, or whitens. | Disputed | The alteration process itself is debated in the literature, and reversion methods are experimental. | No established treatment is described in the sources read. |
| Fresco surface under soot, dust and restorers' animal glue: Colours appear dark and brownish, 'as if through a smoked glass'. | Disputed | The Vatican restorers removed the layers as later accretions. Critics argued that some dark material was Michelangelo's own a secco finishing and was lost with them. | Cleaning by the Vatican Museums' own restorers between 1980 and 1994, as described by the chief restorer. |
| Fresco surface with soluble salts: White efflorescence over the painted surface. | Partly reversible | The veil recurs while water keeps entering; historically it was masked rather than removed. | Earlier restorers hid the efflorescence under glue coatings, which then darkened — one treatment creating the next problem. |
| Lead white in watercolour, gouache and drawing highlights: White highlights turn grey, brown or black. | Partly reversible | Chemical conversion of the darkened pigment has been practised since the nineteenth century, but a review weighs it against changing ethics and its effect on the paper. | Conversion treatment by paper conservators is documented and critically reviewed; it is not described here. |
| Prussian blue and Prussian green watercolour: Blue fades; partial return of colour in air in the dark. | Partly reversible | Recovery on re-exposure to air is incomplete and depends on the sample. | Tate's researchers propose about 5% oxygen, rather than none, for sealed 'anoxic' frames that may contain Prussian blue. |
| Prussian blue dyed on textile: Blue fades; recovers wholly or partly in air. | Partly reversible | After two weeks back in air some samples had fully recovered and others had not; aged historic dye may recover less. | The fumigation study recommends that anoxic treatment not be used on historic textiles dyed with Prussian blue. |
| Scheele's green and emerald green (copper–arsenic greens): Loss or alteration of the green. | Disputed | The mechanism itself is under revision, so no verdict on recovery is warranted. | None reported; identification is the first difficulty. |
| Prussian blue image of a cyanotype: Blue image pales in light and regains colour in the dark. | Partly reversible | Short-term bleaching recovers; for the pigment in paint, long exposure produced loss that did not. | None reported in the source read. |
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
ConventionStrong evidence
ICOM-CC's 2008 terminology defines conservation as embracing preventive conservation, remedial conservation and restoration, distinguished by their aims; preventive measures are indirect and do not modify the appearance of the item.
Caveat: A terminology resolution, not a binding code of practice.
ConventionModerate evidence· United States
The AIC Guidelines for Practice state that compensation for loss should be documented, detectable by common examination methods and reversible, and should not falsely modify the known aesthetic, conceptual and physical characteristics of the cultural property, especially by removing or obscuring original material.
Caveat: A professional body's guideline for its members, in the United States.
StandardStrong evidence
The Venice Charter (1964) states that restoration must stop at the point where conjecture begins (Article 9) and that replacements of missing parts must be distinguishable from the original so that restoration does not falsify the artistic or historic evidence (Article 12).
Caveat: Written for monuments and sites; applied here to colour by analogy.
FactStrong evidence
For Mark Rothko's Harvard Murals, a camera–projector system projects a compensation image onto the faded canvases to restore their colour appearance without altering the paint; the authors compare this 'inpainting with light' with conventional cleaning and inpainting.
Source: Conservation of a room: A treatment proposal for Mark Rothko's Harvard Murals (Studies in Conservation, 2016); Mark Rothko's Harvard Murals (exhibition, 2014–15)
Colourwise analysisModerate evidence
Of 51 recorded colour changes, 37 are classed as irreversible, 10 as partly reversible, 4 as disputed and none as fully reversible.
Based on: Colourwise count of the reversibility verdict recorded for each entry in its colour-degradation dataset. Each verdict is Colourwise's classification of what the cited source reports for the documented case.
Caveat: A count of selected published cases, not a rate for collections at large.
Source: Fading and Colour Change of Prussian Blue: Methods of Manufacture and the Influence of Extenders (National Gallery Technical Bulletin 25, 2004); A review of the phenomenon of lead white darkening and its conversion treatment (Studies in Conservation, 2007)
Colourwise interpretationModerate evidence
Projected-light compensation satisfies the tests of documentation, detectability and reversibility more completely than any treatment of the object, but leaves the question of conjecture open, because the colours projected are an estimate.
Based on: Colourwise's application of the AIC compensation guideline and Venice Charter Article 9 to the Rothko and Hampton Court projections as their sources describe them.
Source: Code of Ethics and Guidelines for Practice; International Charter for the Conservation and Restoration of Monuments and Sites (The Venice Charter, 1964)
Reviewed 6 October 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.