Short answer
Colourwise tabulated 51 documented colour changes by colourant class and environmental driver. Light is the driver named most often (23 records) but in a minority of them; humidity or water is named in 17, and 28 records (55%) involve no light. 37 changes (73%) are recorded as irreversible and none as fully reversible. The matrix counts what has been studied and published, so it shows where evidence exists, not how common a change is.
Each row of the underlying dataset is one documented way that a named colourant changes in a class of object, with the mechanism, the drivers, the observed change and a reversibility verdict taken from the cited study. Colourants are grouped into 14 classes and drivers into nine. The matrix cell for a class and a driver is the number of records of that class that name that driver. A record naming two drivers appears in both columns, so cells add up to more than the 51 records — 30 records name more than one. Two shares are derived: the irreversible share, which is records classed irreversible divided by all records, and the non-light share, which is records with no light driver divided by all records. Nothing is weighted by severity, by number of objects affected, or by strength of evidence.
Light and ultraviolet appear in 23 records, humidity or water in 17, gaseous pollutants in 10, heat in 9, adjacent materials such as binder, support or varnish in 9, acidity or alkalinity in 7, oxygen in 6, chlorides and salts in 6 and low oxygen in 4. The spread is the finding. A collection managed for light alone has addressed fewer than half the documented routes to colour change. Humidity's high count reflects its two roles — reagent in hydrolysis and carrier for salts and soluble degradation products — and the adjacent-materials column is a reminder that some pigments are altered by the paint they are in, with no environmental trigger to control. Low oxygen is the only driver that is also a protective measure for most other materials.
Read across, the classes differ in breadth. The binders, varnishes and coatings class has the widest spread, with 7 of the nine drivers documented, because coatings and binders sit at the surface and meet everything. Copper pigments are the broadest of the true pigments, with 7: they brown in oil and resin, go green with chloride, blacken with heat or alkali and, as verdigris, fade faster without oxygen. Prussian blue is the one class with no record classed as irreversible — every entry is partly reversible, because the reduced pigment regains some colour in air — and the only one where low oxygen is a leading driver. Natural dyes and lakes have the most records and every one is irreversible. Classes with a single record, such as inks or biological colour, are present so that the gap is visible, not because one record characterises them.
The counts follow research attention. Painters whose work has been studied at synchrotrons are over-represented: 4 of the 29 records with a named object concern Van Gogh. Whole categories — dyed leather, lacquer, enamels, most non-European traditions — are thin or absent because Colourwise found no source it could read and verify, not because they are stable. A blank cell means no record. Verdicts on reversibility apply to the case as published; several rest on abstracts alone, and each record carries its own caveat. Drivers are coded as the source names them, so a study that tested only light can only yield a light record. And selection is Colourwise's: another compiler working from the same literature would draw the class boundaries differently and get different cell counts.
The matrix is a reading aid for the evidence, and it supports three modest uses. It tells a reader which conditions to ask about for a given class of material: for a copper green, the binder and any source of chloride; for Prussian blue, whether a sealed or oxygen-free enclosure is involved; for a plastic, temperature and humidity before light. It shows where an 'irreversible' label is well evidenced and where the literature is still arguing. And it points to the object-level pages, where each record's source and caveat are given in full. It is not a risk assessment, it assigns no probability to any object, and it does not indicate treatment. Assessing a particular object's condition and deciding what, if anything, to do about it is the work of a trained conservator.
| Colourant class | Records | Light and UV | Humidity and water | Heat | Gaseous pollutants | Chlorides and salts | Oxygen | Low oxygen (anoxia) | Acidity or alkalinity | Adjacent materials (binder, support, varnish) | Irreversible | Partly reversible or disputed |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Lead pigments (lead white, red lead, lead-tin yellow) | 5 | 2 | 1 | — | 2 | 1 | — | — | 2 | 1 | 2 | 3 |
| Chromate yellows (chrome yellow, zinc yellow) | 2 | 2 | 1 | — | 1 | — | — | — | — | — | 1 | 1 |
| Sulfide pigments (vermilion, cadmium yellow, orpiment, realgar) | 5 | 3 | 1 | — | 1 | 3 | 1 | — | — | — | 4 | 1 |
| Copper pigments (verdigris, emerald green, azurite) | 6 | 1 | 3 | 1 | — | 1 | — | 1 | 2 | 3 | 5 | 1 |
| Prussian blue | 4 | 4 | — | — | — | — | — | 3 | — | 2 | 0 | 4 |
| Glass and mineral colours (smalt, ochre, window glass) | 3 | — | 1 | 1 | — | — | 1 | — | — | 1 | 3 | 0 |
| Natural dyes and lake pigments | 8 | 4 | 1 | — | 3 | — | — | — | 1 | — | 8 | 0 |
| Synthetic dyes and organic pigments | 2 | 2 | — | — | — | — | — | — | — | — | 2 | 0 |
| Writing and drawing inks | 1 | — | 1 | 1 | 1 | — | — | — | 1 | — | 1 | 0 |
| Binders, varnishes and surface coatings | 7 | 2 | 4 | 2 | 1 | 1 | 2 | — | — | 2 | 3 | 4 |
| Plastics | 3 | 1 | 2 | 2 | — | — | 1 | — | 1 | — | 3 | 0 |
| Photographic dyes | 3 | 1 | 2 | 2 | — | — | 1 | — | — | — | 3 | 0 |
| Metal threads and leaf | 1 | — | — | — | 1 | — | — | — | — | — | 1 | 0 |
| Biological colour | 1 | 1 | — | — | — | — | — | — | — | — | 1 | 0 |
| Object class | Records | With a named object | Light among drivers | Not irreversible |
|---|---|---|---|---|
| Easel paintings | 18 | 14 | 12 | 4 |
| Wall paintings, sculpture and architectural surfaces | 10 | 8 | 1 | 5 |
| Manuscripts and works on paper | 8 | 3 | 4 | 2 |
| Textiles | 5 | 2 | 2 | 1 |
| Historic interiors and wallpaper | 2 | 0 | 0 | 1 |
| Photographs and film | 4 | 1 | 2 | 1 |
| Plastics | 3 | 1 | 1 | 0 |
| Natural-history specimens | 1 | 0 | 1 | 0 |
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.
Colourwise analysisModerate evidence
Across 51 degradation records, light is named as a driver in 23, humidity or water in 17, gaseous pollutants in 10 and heat in 9; 28 records (55%) name no light.
Based on: Colourwise count over its colour-degradation dataset: cell(c, d) = number of records with colourant class c whose driver list includes d; non-light share = records without a light driver ÷ all records.
Method: For each colourant class c and driver d, cell(c, d) is the number of degradation records whose colourantClass is c and whose drivers include d. A record with several drivers is counted once in each of its driver columns, so a row's cells can sum to more than its record count. Irreversible share = records of the class with verdict 'irreversible' ÷ records of the class. Non-light share = records with no 'light' driver ÷ all records.
Caveat: Counts published and selected cases, not prevalence in collections.
Source: Large-scale assessment of light-induced color change in air and anoxic environments (Studies in Conservation, 2012); The Blackening of Vermilion: An Analytical Study of the Process in Paintings (National Gallery Technical Bulletin 23, 2002)
Colourwise analysisModerate evidence
37 of 51 recorded changes (73%) are classed irreversible, 10 partly reversible, 4 disputed and none fully reversible.
Based on: Colourwise count of the reversibility verdict recorded for each dataset entry; irreversible share = irreversible records ÷ all records.
Caveat: Each verdict classifies one documented case and may not transfer to another object.
Source: Fading and Colour Change of Prussian Blue: Methods of Manufacture and the Influence of Extenders (National Gallery Technical Bulletin 25, 2004); Blackening of Pompeian cinnabar paintings: X-ray microspectroscopy analysis (Analytical Chemistry, 2006)
Colourwise analysisModerate evidence
30 of 51 records name more than one driver, and Prussian blue is the only colourant class in which no record is classed irreversible.
Based on: Colourwise count over its colour-degradation dataset of records with two or more drivers, and of irreversible verdicts per colourant class.
Caveat: The Prussian blue class has four records.
Source: Photochemical colour change for traditional watercolour pigments in low oxygen levels (Studies in Conservation, 2015); The effect of insect fumigation by anoxia on textiles dyed with Prussian blue (Studies in Conservation, 2004)
Colourwise interpretationModerate evidence
The distribution of records reflects research attention and source availability more than the real frequency of colour change, so the matrix is evidence of where knowledge exists and should not be read as a ranking of risk.
Based on: Colourwise's assessment of its own dataset: 4 of 29 records with a named object concern one painter, and several object classes have one or two records.
Source: Degradation process of lead chromate in paintings by Vincent van Gogh … 2. Original paint layer samples (Analytical Chemistry, 2011); Van Gogh's Irises and Roses: the contribution of chemical analyses and imaging to the assessment of color changes in the red lake pigments (Heritage Science, 2017)
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.