How Paintings Are Restored

What Art Restoration Means in a Studio

ICOM-CC’s 2008 terminology resolution, adopted in New Delhi, splits conservation into three aims. Preventive conservation acts on the surroundings: light, humidity, packing, pests. Remedial conservation acts on an object that is still deteriorating: consolidating flaking paint, reinforcing a torn canvas. Restoration acts on a stable object to make it understandable again: retouching a loss, filling a hole, reducing a yellow varnish that hides the design. Varnish removal can serve both remedial and restorative aims at once. English search language collapses those distinctions. In the studio they stay separate, because the decision to change appearance is not the same as the decision to stop decay.

The American Institute for Conservation (AIC) Code of Ethics adds a further brake. Conservators select methods and materials that, to the best of current knowledge, do not block future examination, scientific investigation or treatment. Compensation for loss must be documented, detectable by common examination methods, and reversible. It must not obscure original material. Icon’s Ethical Guidance, adopted by the UK Institute of Conservation in 2020, tells members to investigate the option of no action first, to keep interventions detectable, and to leave treatments as reversible as possible. ICOM-CC’s 1984 definition of the conservator-restorer already insisted that the worker does not create a new cultural object. Reconstruction of what no longer exists belongs to craft. The studio’s job is to keep the original readable.

A paintings conservator therefore spends more time looking than painting. The National Gallery of Art in Washington states that treatment follows AIC guidelines and that conservators work with curators and scientists on attribution, prior treatment and the artist’s procedure. The Met’s Sherman Fairchild Paintings Conservation Center, opened in 1980, put studio, panel work, lining, photography and analysis in one duplex. The V&A’s paintings section reports that about 90 per cent of its work is driven by exhibitions, loans and gallery projects, and that the default is to wait until a more invasive treatment is necessary.

Materials and Chemistry Conservators Have to Read

A painting is a stack of materials with different ages, solubilities and mechanical strengths. Treatment that ignores that stack will dissolve a glaze, stain a ground, or glue a canvas into a sandwich that later conservators cannot undo. Before anyone wets a swab, the conservator has to know what the artist put down, what later hands added, and how those layers now respond to water, heat and organic solvents. The chemistry is a map of risk, not a recipe list. Easel paintings share a rough architecture: a support carries a ground; paint sits on the ground as pigment in a binder; a varnish, if present, sits on top as a sacrificial film. Dirt, nicotine, overpaint and earlier restorations sit on or in those films. The conservator’s first chemical task is to tell those layers apart.

Support, ground and paint film

Canvas supports are linen or cotton, stretched on a wooden stretcher or strainer. Size (animal glue) and a ground (chalk and glue, lead white in oil, acrylic gesso) prepare the cloth. Tears, holes and slack tension are structural problems. So is a lining: an extra fabric adhered to the reverse, in older practice with glue-paste or wax-resin, later with heat-seal adhesives such as Beva 371, which Gustav Berger published for conservation lining in the 1970s. Niccola Shearman, writing on the V&A in the Getty’s Conserving Canvas, notes that in twenty-five years she had met one canvas she judged needed lining. Current practice prefers local tear mends over a full lining that changes the mechanical behaviour of the whole picture.

Panel paintings add wood movement. Boards shrink and swell across the grain. Cradles glued to the reverse in the nineteenth century often caused new splits. Conservators now favour flexible battens or climate control rather than rigid restraint. Reverse-glass paintings fail by paint detaching from the glass; a V&A Gainsborough treatment chose Laropal A81 after ageing tests, because the adhesive would sit between paint and glass and could not be allowed to yellow.

The paint film is pigment plus binder. Oil paint that is underbound, or that contains certain lakes, can be more soluble than the varnish above it. Tempera is water-sensitive. Acrylic dispersion paint can swell in water and in some solvents that leave aged oil untouched. The Getty Conservation Institute’s Solvent Gels for the Cleaning of Works of Art underlines a point every code already makes: cleaning is irreversible in practice. You cannot put dissolved varnish, or dissolved original glaze, back on the surface.

Varnish, dirt and later additions

Natural resin varnishes (dammar, mastic) saturate colour and even gloss. They also oxidise. Dammar yellows and can crack. Mastic cross-links as it ages, so the solvents that dissolved it when fresh may no longer remove it without risk to the paint. Synthetic coatings from the mid-twentieth century onwards include ketone resins (MS2A), hydrocarbon resins (Regalrez 1094), urea-aldehyde resins (Laropal A81) and acrylics (Paraloid B-72). Low-molecular-weight synthetics were adopted because they saturate more like dammar while remaining soluble after ageing. Paraloid B-72 is a tougher acrylic film, useful as an isolating layer or consolidant, less saturating as a final varnish on some dark oil paintings. The National Gallery of Art re-treated Rembrandt’s The Mill in 2001 because a varnish applied in 1979 had stopped saturating the dark paint.

Dirt is a separate layer: soot, grease, nicotine, fly specks.

Overpaint is later paint covering original, sometimes to hide damage, sometimes to change a face. Retouching confined to losses is the ethical form of compensation. Overpaint that covers intact original is a candidate for reduction, if it can be removed without taking the artist’s paint with it. Ultraviolet fluorescence helps here: many natural varnishes fluoresce greenish; many later retouches appear dark. Infrared reflectography and X-radiography belong to a sibling guide on imaging; in this studio sequence they are tools for finding pentimenti, old tears and the extent of fill.

Examination and Documentation Before Treatment

AIC’s Guidelines for Practice put examination and written records before treatment. Icon’s 2020 principles say the same: know why an action is proposed, understand the causes of deterioration, and document the action or the decision to take none. The next conservator, twenty or sixty years on, will inherit your solvents, your fills and your guesses. If the file is thin, they will have to rediscover the painting by putting it at risk again.

A full examination starts with the object out of its frame, recto and verso, in ordinary light, raking light and ultraviolet. The stretcher, labels, wax seals, old patches and inscriptions on the reverse are part of the history. Photography in those light modes, plus a written condition report, is the minimum. Many studios add infrared and X-radiography when the question is structural. Sampling is reserved for questions that non-invasive methods cannot answer, because every sample is a loss.

Light, imaging and the microscope

Raking light shows cupping, cleavage and the topography of old fills. Specular light shows abrasion and the sheen of overpaint. Ultraviolet induced fluorescence maps varnish and many retouches. The Met’s treatment of a Velázquez Portrait of a Young Girl used that sequence in public: thick discoloured varnish had flattened the cool palette; after varnish reduction, raking light showed raised, insecure paint that needed consolidation before anyone filled the edges. Larry Keith’s 2008-9 treatment of Leonardo’s Virgin of the Rocks at the National Gallery followed years of analytical study. The 1949 oil-and-mastic varnish had cracked and yellowed until the picture was hard to read. Cleaning went ahead for aesthetic reasons, but only after the layer structure was known.

Infrared reflectography can reveal carbon underdrawing. X-radiography can reveal canvas weave, old tears, pentimenti in lead-white passages, and nails or keys. The NGA’s four-year treatment of Tiepolo’s Bacchus and Ariadne used infrared and cross-sections to recover gold ribbons that later overpaint had hidden. Those methods have limits, which the imaging article in this series will treat at length.

The stereomicroscope is the working instrument. Joanna Dunn, treating a Tintoretto at the NGA, told NPR in 2017 that an aged coating would not dissolve without harming the paint, so she softened it and eased the gelled film off with a scalpel under magnification. That is not a method to copy at home. The Getty’s work on Manet’s The Ragpicker is the cautionary case: a synthetic varnish had bitten into original paint. Full removal would have taken the artist with it. The conservator thinned the coating instead.

Written records, sampling and the decision to stop

AIC requires dated documentation of techniques, materials and their composition, the nature and extent of all alterations, and information revealed during treatment. A report drawn from those records goes to the owner or custodian, with recommendations for later care. Copies travel with the object. A painting sold without its treatment file has lost part of its history.

Sampling is a research tool, not a default. Cross-sections, examined in visible and ultraviolet light and with scanning electron microscopy, show layer order: ground, paint, varnish, overpaint. Gas chromatography-mass spectrometry and infrared spectroscopy can identify binders and coatings. The National Gallery’s Virgin of the Rocks campaign used that suite to clarify surface layers during cleaning. Private studios without a scientist on staff still write down what they can see, what they tested, and what they left.

The hardest documented decision is to stop. AIC states that when nonintervention best serves preservation, the conservator may recommend that no treatment be performed. Icon asks members to investigate no action first. A painting with sensitive glazes or a varnish fused to original paint may be safer with a dirty surface and a good frame than with a bright surface and a thinned original. Preventive measures (the subject of a later article in this series) then carry the object. Restoration can wait. Flaking paint cannot.

The Treatment Sequence: Cleaning, Filling, Retouching, Varnish

Once the file exists and the object is stable enough to handle, treatment follows a conservative order. Loose paint is secured before anyone cleans. Cleaning, if it happens, comes before filling, because you cannot match a colour you have not seen. Fills are brought to level. Retouching sits on the fill, not on original paint. A new varnish saturates the surface and isolates the retouching. Structural work on canvas or panel may interrupt that sequence: a tear is mended before a heavy cleaning campaign, or a panel split is addressed before the picture goes back on the wall. The table below is a studio map, not a checklist you run at home. Solvent choice, gel formulation and adhesive concentration are professional judgements made after tests on that object. Publishing recipes here would invite amateur treatment, which this series refuses. Museum campaigns follow the same order at different speeds: Constable’s The Hay Wain and Rubens’s Het Steen at the National Gallery, and the Met’s Haverman A Vase of Flowers, all ran examination and analysis before retouching.

Cleaning and varnish removal

Cleaning in conservation language includes dirt removal and the reduction of unwanted coatings. Getty authors note that it is one of the few procedures that is irreversible: dissolved material does not return. Tests come first, and "full clean" is a decision, not a default. The National Gallery describes Keith’s Het Steen solvents as a mixture that dissolved the varnish alone. This article will not list recipes or ratios. The professional point is discrimination: the coating must move, the paint must not.

Gels and emulsions exist because free solvent on a swab can flood a lean film. Gels hold the solvent on the surface, limit penetration, and can be cleared. Residue is a research topic of its own; the Getty gel volume is the standard warning. Mechanical reduction under the microscope, as at the NGA, is the option when coating and paint share solubility. None of these methods is a household product with a new name.

Overpaint removal is slower than varnish removal. Conservators leave overpaint in place when removal would take original with it. An NGA intern treatment of a Le Lorrain masquerade scene, summarised by the Washington Conservation Guild, records that choice on a degraded smalt overpaint: the isolating layer was too thin to protect the original, so the overpaint stayed.

Filling, retouching and a new varnish

Fills replace lost ground and paint in volume, not in invention. The Met’s Velázquez sequence is the textbook order: consolidate, apply a thin saturating varnish as an isolating layer, bring edge losses to level with a fill toned to the ground, then retouch so the damage no longer shouts. AIC’s compensation rule is strict: retouching must be detectable by common examination methods, reversible, and confined so that it does not obscure original. Imitative inpainting matches the surrounding colour so the loss disappears at normal viewing distance. Visible reconstruction (Italian tratteggio or rigatino) is a Brandi-school alternative when the loss is large and a fake original would lie.

Retouching media in current easel practice are chosen for reversibility. Gamblin Conservation Colors bind pigment in Laropal A81; the V&A Gainsborough reverse-glass treatment used them on small losses. Oil paint as retouching is a historical habit that later darkens. The isolating varnish beneath the retouching is the safety net: if the colour is wrong, the next conservator can lift the restoration without stripping the artist.

A final varnish saturates the paint, protects it from dirt, and unifies gloss. A matte modern painting may need no varnish. Dammar still has defenders for its optical properties. Synthetics (Regalrez 1094, Laropal A81, MS2A, Paraloid B-72) were adopted to reduce yellowing and to keep future removal in milder solvents. The NGA Mill re-treatment in 2001 is the reminder that saturation can fail even when the resin is synthetic. Varnish is a sacrificial layer. It is meant to be replaceable.

Structural Work: Tears, Linings and Panels

A beautiful surface on a failing support is a short-term result. Canvas tears propagate. Cupping paint will flake if the cloth keeps flexing. Panel splits open in a dry gallery. Structural treatment is remedial conservation in ICOM-CC’s sense: it arrests damage. It also changes the object. Every lining, cradle and transfer is a historical event that later conservators have to live with.

Tear repair in current practice aims to rejoin original fibres where they survive. Thread-by-thread mending, local patches and strip-lining add support without covering the entire reverse. Full lining was standard for generations. Glue-paste linings can flatten impasto, leave stains and shrink. Wax-resin linings impregnate the original canvas, darken some pictures, and resist reversal. Beva 371 and related heat-seal systems were developed to line at lower temperature with a more reversible adhesive. Even those are now used with more hesitation. The Getty Conserving Canvas project exists because lining itself became a cause of deterioration. The V&A’s published stance (lining as a last resort) is the institutional version of that lesson.

Transfers are historical operations, not a current first choice. They destroy evidence of the original support. Panel work now tries to accommodate movement: rigid cradles have given way to flexible systems and to climate specifications. Frames are part of the structure. A rebate that bites the paint will undo a year of studio work. The V&A’s gallery projects treat frames and glazing as conservation, not as decoration.

How the Profession Took Shape

Picture restoration as a trade is older than the word "conservator." Workshop assistants, dealers and specialist cleaners have been removing varnish and adding paint for centuries. Filippo Baldinucci, in the seventeenth century, already named patina as the darkness time lays on pictures, sometimes to their benefit. The modern profession begins when that trade meets laboratories, codes, and the claim that the restorer must not invent.

Italy built an institutional model in 1939: the Istituto Centrale del Restauro in Rome, designed by Giulio Carlo Argan and Cesare Brandi, with Brandi as first director. The institute put art historians, scientists and restorers in one public laboratory. Brandi’s Teoria del restauro (1963) gathered the lessons of that work. Restoration, for Brandi, is a critical act that respects both the aesthetic instance of the work and its historical instance. Patina is not dirt by default. Cleaning that strips glazes in search of a raw original can falsify the object as much as a sugary overpaint.

London supplied the opposing pole in public. During the Second World War the National Gallery cleaned large numbers of pictures. In 1947 it exhibited more than seventy of them as Cleaned Pictures. Helmut Ruhemann, consultant restorer at the Gallery, stood for a practice that trusted chemistry to separate varnish from paint. Titian’s Bacchus and Ariadne became a flashpoint. Brandi, Ernst Gombrich and others argued in The Burlington Magazine that total cleaning could remove original glazes. The Weaver Committee investigated. The controversy, which ran into the 1960s, pushed the Gallery to strengthen its scientific department. Ruhemann’s manual The Cleaning of Paintings: Problems and Potentialities appeared in 1968.

North America absorbed both traditions. John Brealey, a London private restorer, became head of paintings conservation at the Metropolitan Museum of Art in 1975 at the invitation of John Pope-Hennessy, and stayed until a stroke ended his working life in 1989. Brealey taught connoisseurship as the core of treatment: the look is the meaning. AIC’s codes, Icon’s accreditation (Accredited Conservator-Restorer), and ICOM-CC’s 1984 and 2008 texts are the bureaucratic afterlife of those fights. They tell you that you will not be the last person to treat the object, and that your materials must leave that person a way in.

Practitioners Who Set the Standard

Named individuals matter here because students copy habits, not slogans. Ruhemann (1891-1973) trained a generation in London in close, solvent-based cleaning and in the belief that the artist’s colour was recoverable. His archive at the Hamilton Kerr Institute still holds the paper trail of the post-war cleaning disputes, including his clash with Gombrich. Brandi (1906-1988) gave European restoration a theory that still governs Italian training: visible, reversible compensation; respect for time as well as for the first surface.

Brealey (1925-2002) exported a painter’s eye to the Met and ran seminars for museum directors so that the people who commission treatment would know what they were looking at. Paul Coremans, working from Brussels and from the Courtauld orbit, defended the chemistry of cleaning while admitting that aesthetic criteria still judge the result. That double demand (test, then look) is now ordinary.

Present museum practice is a team. Larry Keith’s published campaigns at the National Gallery (Virgin of the Rocks, Het Steen) show the current English institutional style: years of analysis, solvents chosen to spare paint, structural specialists for panels, retouching as a separate campaign, and a Technical Bulletin so the file is public. Jay Krueger and colleagues at the NGA describe varnish as the recurring enemy of colour, and still leave overpaint when removal is unsafe. Niccola Shearman at the V&A treats lining as an exceptional event and loan preparation as the bulk of the job. The mastery on display is restraint: knowing when the swab stops.

How Treatments Age, Fail and Have to Be Undone

Every material in the table has a failure mode. Natural resin varnish yellows, cracks, and can become less soluble. Synthetic varnish can blanch, lose saturation, or pick up dirt in a tacky film. Oil retouching darkens and crosses from fill onto original. Glue-paste linings stain and draw. Wax-resin linings darken and block later adhesive work. Overcleaning cannot be reversed: glazes and scumbles, once gone, are gone. Amateur "restoration" with household solvents, farm oils or mayonnaise sits in that irreversible class. The pending controversies article collects the public cases. The studio lesson is smaller and more frequent: yesterday’s best lining is today’s problem.

Ageing is chemical and mechanical. Oil films continue to cross-link and become more polar; a solvent mixture that was safe in 1970 may be less safe on the same picture in 2026 because the paint has changed. Canvas fibres weaken in acid conditions. Fluctuating humidity opens cleavage. Light fades lakes. A treatment that ignored the cause (a damp wall, a heater under the picture, direct sun) will be repeated. Icon’s principles put underlying causes of deterioration on the same plane as the bench work. AIC’s preventive clause does the same: environment, handling, packing and display are part of the treatment.

Re-treatment is the expected future. AIC’s continuum-of-care sentence is the most honest line in the code: the conservator will seldom be the last person entrusted with the object. Isolating varnishes, detectable fills and soluble retouching exist so that re-treatment is a campaign, not an archaeological excavation. The NGA’s Mill (treated 1977-79 amid a public fight over whether removing golden varnish would destroy the picture’s mood, then varnished again, then re-varnished in 2001) is a single object with three documented campaigns. That is success in conservation terms. The picture remained treatable.

What Painting Restoration Costs

Cost follows time, materials and risk, not the market value of the painting. The Williamstown Art Conservation Center and Plowden & Smith in London both state that the invoice tracks hours and materials: a child’s drawing and a Rubens take the same standard of care. Value may change how much documentation a client wants. It should not change the chemistry.

Published US private-practice illustrations sit in a wide band. Peter Himmelstein, a paintings conservator at Appelbaum & Himmelstein, has been quoted putting a small painting with an average amount of work in the region of 800 to 1,000 dollars, and a larger painting with serious damage in the region of 10,000 to 15,000 dollars. Treat those figures as order-of-magnitude hints from interviews, not as a 2026 tariff. Examination and a written proposal are the real quote.

Museum treatments are staff time, not a public price list. The National Gallery’s Virgin of the Rocks took about eighteen months of treatment after years of research. The NGA’s Tiepolo took four years. A private studio restoring a modest Victorian portrait with a yellow varnish and one tear may finish in weeks. A lining or a panel with active splits will not. Reputable studios ask for photographs of front and back, then an in-person examination, then a written scope. Icon’s Conservation Register and AIC’s Find a Conservator exist so that you hire a professional bound by those codes.

Ethics: Reversibility, Detectability and Doing Nothing

Ethics in this field is a set of operational limits, not a mood. The object is unique. Treatment is a physical change. Codes exist because taste is a poor control on solvents. If you cannot document, reverse and detect what you are about to do, you are not restoring the painting. You are altering it without a record.

Icon’s 2020 list puts no action first, least means next, then detectability, then reversibility as far as possible, then documentation. AIC’s compensation paragraph requires all three for fills and retouches: recorded, visible to ordinary examination, removable. ICOM-CC’s restoration definition adds respect for original material. Famous disasters have their own article in this series.

Reversibility as a practical limit

Barbara Appelbaum’s 1987 JAIC paper on the principle of reversibility remains the honest technical account. No treatment is reversible at the molecular level. Traces of modern materials can interfere with later analysis. In practice, conservators ask whether the added material can be removed without endangering the object; whether removal is a dissolution or a mechanical lift; how much time, cost and risk that removal would take; and whether the object can still be re-treated if the addition has to stay. Wax-resin lining fails those tests. A Laropal retouch on an isolating varnish passes them. Cleaning, as the Getty gel volume says, is irreversible by nature. That is why cleaning is the most argued, most tested, and most documented step.

"As reversible as possible" is Icon’s phrasing. Consolidation adhesives that must stay in the paint film to do their job are a known compromise. So are some panel joins. The obligation is to avoid materials that become intractable, and techniques whose results cannot be undone if that becomes desirable (the older AIC language). Choosing a resin because it is familiar, or because it makes the picture look sale-room bright, fails that test if the film will cross-link into the original.

Detectability, documentation and the owner’s file

Retouching that cannot be seen under ultraviolet, magnification or gentle solvent tests is a historical lie. AIC wants compensation detectable by common examination methods so that a future conservator can map your work without destroying it. Italian hatched reconstruction makes the repair visible to the naked eye at close range. Imitative inpainting relies on the file and on UV. Both are ethical if the original remains under the restoration, not under a blanket of new paint. Overpaint that "improves" a face or hides a pentimento modifies the known characteristics of the work. Codes forbid that except where the later paint is itself historic and significant.

Documentation is the ethical act that outlives the varnish. Photographs before, during and after; a list of materials with trade names; a map of fills; a note of what you did not do. AIC says copies go to the owner, who should keep them with the object. If you own the work, ask for the report in writing. If you are commissioning treatment, ask whether the studio follows AIC, Icon or an equivalent code, and whether the retouching is reversible.

FAQ

Search results for "art restoration" mix local-studio ads, Wikipedia, YouTube cleanings and career questions. The questions below belong on a process page. Training and salaries are covered in the live guide on how to become an art conservator. Owner packing and dusting are reserved for caring-for-art. Viral "restoration fails" are reserved for the pending controversy article.

A qualified paintings conservator will still begin with the same two moves: look at the reverse as well as the front, and write down what they propose before they wet a swab. If a quote arrives without that examination, treat it as marketing.

How much does it cost to get a painting restored?

Invoices follow hours and materials, not market value. Published US private-practice examples put a small painting with modest work around 800 to 1,000 dollars and a larger damaged canvas in the low-to-mid tens of thousands. London firms such as Plowden & Smith state the same time-based rule. The usable number is a written estimate after the conservator has seen the reverse as well as the front.

Can I restore a painting myself?

No. Cleaning is irreversible. Household solvents, oils and hardware-shop "picture restorers" do not discriminate between dirt, varnish and original glaze. Flaking paint needs consolidation before any surface work. Limit yourself to dusting the frame, improving storage, and a professional examination. The caring-for-art guides cover that owner-level stop line.

What is the difference between art restoration and art conservation?

ICOM-CC (2008) uses conservation as the umbrella. Preventive conservation manages the environment. Remedial conservation stabilises an object that is still deteriorating. Restoration changes the appearance of a stable object so it can be understood: retouching, filling, reducing a discoloured varnish. English search language uses "art restoration" for the whole field. In a studio, restoration is the visible subset, and it is optional in a way that consolidation of flaking paint is not.

How long does painting restoration take?

A straightforward varnish reduction on a stable small canvas may take days or weeks of bench time, plus drying. Structural repair, cupped paint or a large loss map take months. The National Gallery’s Virgin of the Rocks treatment ran about eighteen months after years of research. The NGA’s Tiepolo Bacchus and Ariadne ran four years. Anyone who promises a museum picture back in a weekend is not describing conservation.

Why do conservators remove varnish?

They remove or thin varnish when the coating has yellowed, cracked or dirtied to the point that it hides the design, and when tests show that the coating can move without taking original paint. Rubens’s Het Steen and Constable’s The Hay Wain at the National Gallery are published examples: seventy-five to eighty years of discoloured varnish had shifted colour and buried detail. They leave or thin a coating when it is fused to original, as with Manet’s The Ragpicker at the Getty.

How do I find a qualified paintings conservator?

Use AIC’s Find a Conservator in the United States and Icon’s Conservation Register in the UK. Look for paintings (easel) as the specialism, and for AIC Professional Associate or Fellow status, or Icon ACR. Ask for a written examination report, a treatment proposal, and a statement of materials. ICOM-CC’s 1984 definition is the test: the conservator-restorer does not create a new cultural object.

Is restoration reversible?

Fills, retouching and modern varnishes are designed to be as removable as current materials allow. Cleaning is not. Old wax-resin and glue-paste linings often were not designed for later removal. Appelbaum’s point still stands: molecular reversibility is a myth; practical retreatability is the real standard. Many pictures need preventive care more than a campaign. A yellow varnish that still saturates, on a stable canvas in a sound frame, may be left. Flaking, active tears and a coating that is pulling paint off are not "wait and see" problems.