Short answer
They combine three kinds of evidence: colour that survived where light never reached, chemical maps showing which pigment is where and how much, and fresh samples of the same materials made to historical recipes. A model then calculates what the unaltered mixture would have looked like, and the result is shown on a screen, by projected light or on a replica — never by repainting the object. Every published reconstruction states its own gaps, and those gaps are usually the very pigments that changed most.
The best witness is a part of the object the damage missed. On a painting it is the strip under the frame rebate or a covered edge; the study of Van Gogh's Irises and Roses measured colour on protected edges and in paint cross-sections, where the interior of a paint layer keeps colour the surface has lost. On a tapestry it is the back: for Henry VIII's Abraham tapestry at Hampton Court, researchers spent three months measuring yarns on the reverse, where the thick fabric and lining had slowed fading. On stone it is a trace of pigment caught in a fold of carving. Documents help as well — the artist's letters, early reviews and early reproductions, all used in the Van Gogh studies. Each witness has limits: an edge has aged too, only more slowly, and an early colour photograph has faded in its own way.
Scanning methods turn a painting into a set of maps. Macro X-ray fluorescence records elements point by point, so lead, mercury, chromium or copper can be followed across the whole surface; reflectance imaging spectroscopy records a spectrum at every pixel and separates pigments that share a colour; X-ray diffraction mapping distinguished the lightfast from the light-sensitive chrome yellow across Van Gogh's Sunflowers without a sample being taken. For Vermeer's Girl with a Pearl Earring, infrared reflectography, optical coherence tomography and 3D microscopy revealed a signature and folds of fabric in a background the eye sees as flat black. The Metropolitan Museum's team noted the limit of the non-invasive route: where paints are layered and mixed, microsamples were still needed to interpret the maps.
Organic dyes are harder: they are present in tiny amounts and carry no heavy element, so they are identified by chromatography on a sample or inferred from fluorescence.
A map says what was used, not what it looked like when new. For that, researchers remake the materials. The Van Gogh Museum's project prepared paints for every pigment in Field with Irises near Arles following historical sources on raw materials and manufacture, then measured their optical properties. The Manchester team recreated sixteenth-century dyeing with weld, old fustic, woad and madder and artificially aged the results to learn how each dye shifts. For Seurat's La Grande Jatte, paint-outs of unaltered pigments stood in for the darkened zinc yellow. The measured spectra feed a mixing model — Kubelka–Munk theory, which predicts the reflectance of a pigment mixture from the absorption and scattering of its components — and the model returns a spectrum, and so a colour, for each point on the map.
There are three ways to present a reconstruction, and all of them leave the object as it is. A digital image, as for the Seurat and Van Gogh paintings, is the cheapest and the easiest to revise. Projected light adds the missing colour optically on the object itself: Harvard's compensation image for the Rothko murals was computed from digitally restored photographs taken when the paintings were new, and the Hampton Court projection lit some two million separate areas of the tapestry. A physical replica, as in the painted casts of Greek and Roman sculpture, lets the result be seen in the round and in real light. The choice affects how the claim reads. A replica looks finished and certain; a projection that is switched off each afternoon reminds the viewer that it is an estimate.
The published studies are candid about this. The Field with Irises team wrote that technical-scientific data 'was not sufficient to resolve several issues' — the sky, and how much the red lakes originally contributed — and that those were settled by discussion within an interdisciplinary team. The Irises and Roses authors say their simulations visualise the original 'to a certain extent'. The Seurat paper is titled a simulation, and must also estimate a general darkening of the whole picture. The pattern is consistent: the pigment that altered most is the one whose original concentration and colour are least recoverable, because the evidence went with it. Reference photographs fade; remade paints follow recipes that varied between makers; the result is shown under a modern light source the painter never used. A reconstruction is best read as a bounded hypothesis.
| Object | What changed in appearance | Method | Rests on | What is disputed or uncertain |
|---|---|---|---|---|
| Mark Rothko, Harvard Murals (five canvases, 1962) (Harvard Art Museums) | Canvases faded by daylight appear in their original deep colours while the projectors run, and faded when they are switched off. | A camera–projector system projects a calculated compensation image onto each canvas. | Contemporary Ektachrome transparencies, themselves faded and digitally restored. | The museum presented it explicitly to encourage 'study and debate of the technology'. The target colours rest on restored photographs, and the authors set the approach against conventional cleaning and inpainting rather than claiming it replaces them. |
| The Oath and Departure of Eliezer, from the Abraham tapestries of Henry VIII (Hampton Court Palace) | A faded sixteenth-century tapestry was shown, at timed intervals, in bright estimated original colours. | Calibrated light projected onto about two million sections of the surface, each tuned to shift a yarn to its estimated colour. | Yarns re-dyed with weld, old fustic, woad and madder to period methods and artificially aged; colour measured on the less-faded reverse. | Known to Colourwise from a university press account. The original colours are estimates from re-dyeing and from the reverse, which has itself aged. |
| Georges Seurat, A Sunday on La Grande Jatte—1884 (Art Institute of Chicago) | Ochre, olive and reddish-brown strokes are shown as the yellow, green-yellow and orange they are calculated to have been. | Spectral reflectance measured on the painting and on paint-outs of unaltered colours, modelled with Kubelka–Munk theory and applied to a colour-managed image. | In-situ spectral measurements and fresh paint-outs of comparable pigments. | The authors title it 'a simulation'. It corrects the zinc yellow only against a general darkening of the whole painting that it also has to estimate. |
| Vincent van Gogh, Field with Irises near Arles (Van Gogh Museum, Amsterdam) | Purplish-blue irises are reconstructed as warm purple, white spots as pink, and a narrowed range of yellows as wider and brighter. | Pigment concentration maps combined with the measured optical properties of paints remade from historical recipes. | Pigment maps, physical paint reconstructions and Van Gogh's own descriptions in letters. | The authors state that technical data could not resolve the sky or the original contribution of the red lakes, and that those were settled by discussion within the team. |
| Vincent van Gogh, Irises and Roses (The Metropolitan Museum of Art, New York) | Faded red lakes are restored on screen, changing how the two still lifes relate to each other. | Digital colour simulation from MA-XRF maps, reflectance imaging and microsamples. | Colour measured on cross-sections and protected edges, the artist's descriptions, early reviews and reproductions. | The authors say the simulations visualise the original appearance only 'to a certain extent'; layered and mixed paints defeat non-invasive methods alone. |
| Abraham Mignon, seventeenth-century flower still life (As named in the cited study) | A yellow rose that now looks flat is shown with the light-to-shadow modelling the painter built. | Macro- and microscale chemical imaging of the surviving and altered pigments. | Element and compound maps of orpiment, its degradation products and a faded yellow lake. | A single-painting case study; the reconstruction is an image, not a treatment. |
| Johannes Vermeer, Girl with a Pearl Earring (c. 1665) (Mauritshuis, The Hague) | A background read as flat black is shown to have been a green glaze with the suggestion of a folded curtain. | Cross-sections, chromatography, infrared reflectography, optical coherence tomography, MA-XRF and 3D digital microscopy. | Indigo and weld identified in the glaze over a charcoal-black underlayer. | The curtain is described as a 'suggestion' of folded fabric; how green and how legible it once was is not quantified. |
| The Parthenon Sculptures (British Museum, London) | Marble long treated as white is shown to carry traces of Egyptian blue, used for figurative designs on carved drapery, and of a purple colourant. | Close examination and archaeometric imaging of the surfaces. | Surviving pigment traces and tool-finished surface textures. | Traces show that colour was applied and where; they do not give the full original scheme, and the authors call for reconsideration rather than offering a reconstruction. |
| Greek and Roman marble sculpture (more than 30 painted reconstructions) (Liebieghaus Skulpturensammlung, Frankfurt) | Full-size painted copies stand beside the bare originals, showing patterned garments in strong colour. | Physical replicas painted on the basis of pigment traces found on the originals. | Surviving traces of polychromy and ancient literary sources. | The museum page presents the reconstructions as research results and does not discuss their uncertainty. A replica must commit to a colour everywhere, including where no trace survives. |
| Kind of intervention | Cases | Original material altered | Examples |
|---|---|---|---|
| Treatment of the object | 5 | 5 | Michelangelo; Paintings shown in the 1947 exhibition of cleaned pictures |
| Compensation with projected light | 2 | 0 | Mark Rothko; The Oath and Departure of Eliezer |
| Digital reconstruction | 4 | 0 | Georges Seurat; Vincent van Gogh |
| Physical reconstruction | 1 | 0 | Greek and Roman marble sculpture |
| Analytical evidence of a lost appearance | 2 | 0 | Johannes Vermeer; The Parthenon Sculptures |
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.
FactStrong evidence
The digital reconstruction of Van Gogh's Field with Irises near Arles shows that, through discolouring of red lake pigments, the irises have changed from warm purple to purplish blue and many pink spots to white; the authors state that technical-scientific data was not sufficient to resolve several issues, which were decided by discussion within the team.
Method: Pigment concentration maps combined with optical properties of physical paint reconstructions.
Caveat: One painting; the reconstruction rests on remade paints.
FactStrong evidence
For Seurat's La Grande Jatte, the original appearance was simulated from in-situ spectral reflectance measurements, paint-outs of comparable unaltered colours and Kubelka–Munk turbid-media theory, applied to a colour-managed image.
Caveat: Described by its authors as a simulation.
FactModerate evidence
For the Hampton Court tapestry The Oath and Departure of Eliezer, original colours were estimated from yarns re-dyed by sixteenth-century methods and artificially aged, and from measurements on the less-faded reverse, then shown by calibrated light projected onto about two million sections.
Caveat: From a university press release, not a peer-reviewed account.
Source: Scientists 'virtually restore' 16th century tapestry at Hampton Court Palace
FactStrong evidence
Imaging of Vermeer's Girl with a Pearl Earring showed that the dark background was a glaze of indigo and weld over a black underlayer, whose deterioration hides the artist's signature and a suggestion of folded fabric.
Colourwise analysisModerate evidence
Of the 14 documented cases compiled here, 9 recover or show an earlier appearance without altering the object and 5 are treatments of the object itself.
Based on: Colourwise count over its before-and-after case set, classing each case by whether the cited source reports any change to original material.
Caveat: A selection of published cases chosen to illustrate methods, not a survey of practice.
Source: Digitally reconstructing Van Gogh's Field with Irises near Arles. Part 3: Determining the original colors (Color Research & Application, 2018); Conservation of a room: A treatment proposal for Mark Rothko's Harvard Murals (Studies in Conservation, 2016)
Colourwise interpretationModerate evidence
In colour reconstruction the largest uncertainty attaches to the colourants that changed most, because their original amount and colour are what the evidence no longer holds.
Based on: Colourwise's reading of the limitations the Van Gogh reconstruction studies state for themselves, both of which concern the faded red lakes.
Source: Digitally reconstructing Van Gogh's Field with Irises near Arles. Part 3: Determining the original colors (Color Research & Application, 2018); 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.