How to Look at a Painting in a Gallery

Find the light before you look at the picture

Gallery lighting is a conservation decision before it is a display decision, and knowing the numbers changes how you use a room. Standard guidance caps the most sensitive material, meaning works on paper, photographs and textiles, at around 50 lux; less sensitive material including oil paintings, wood and leather at around 150 lux; and the least sensitive material such as metal, ceramic, stone and glass at around 300 lux. Ultraviolet is typically held to 75 microwatts per lumen or well below. Damage is cumulative, calculated as light level multiplied by hours of exposure, so institutions also ration display time: something in the region of three months in every five years for the most sensitive class, three months in every two years for oils.

Two things follow for the visitor. An oil painting is being lit at roughly a third of comfortable reading light and a drawing at a sixth of that, so your eyes need several minutes in the room before any judgement about colour is worth making. And the darkness in the drawings gallery is a rationing calculation, not an atmosphere.

You cannot move the lamps, so move yourself. Step to one side until the spotlight falls nearly parallel to the picture plane, then look along the surface rather than at it. This is a crude version of raking light, the technique conservators use with a lamp at an oblique angle to read surface topography; the Getty's published treatment of Artemisia Gentileschi's Self-Portrait as Saint Catherine of Alexandria describes examining the surface under acute raking light to locate raised areas before lining. What comes up at that angle is everything the frontal view suppresses: raised and cupped cracks, dents and draws in a canvas, flattened impasto, the ghost of a patch on the reverse, and above all uneven gloss.

Varnish is the first thing you are seeing, and it is not the artist's

Traditional varnishes are natural resins, and they yellow, cloud and crack as they age. The optical consequence is systematic rather than random. Blues shift towards green, whites towards cream, and the darkest passages read brown while the interval between them and the mid-tones collapses. A picture's sense of depth flattens. Any statement about a painter's colour made through a hundred-year-old varnish is at least partly a statement about the varnish, and the safest habit is to say so in the sentence.

Bloom is the other common film defect: a milky, slightly blue haze across the surface, associated with moisture acting on a degraded varnish layer. It looks like a fault in your own vision until you move and it stays put.

Uneven gloss under the raking angle is the single most useful observation available to a visitor, because retouchings and fills rarely match the surrounding surface exactly. What matters is less that they are there, since almost every old picture has them, than what shape they make. A scatter of small spots through a sky reads as loss from flaking. A hard-edged linear zone reads as a repaired tear. A broad, even matt area reads as abrasion that has been glazed back, which typically means thin original paint has been lost and the passage in front of you is largely a restorer's reconstruction.

The point of noticing is not to catch a museum out. It is that condition governs what you can legitimately say about handling. A passage that looks weak, coarse or oddly flat may not be by anybody in particular, and a connoisseurial argument built on it will not survive contact with the technical file.

Cracks record materials and structure, not a certificate of age

There are two basic mechanisms. Drying cracks form as a paint film shrinks while it dries over a dimensionally stable substrate, and they tend to appear in particular passages, typically the rich, slow-drying browns and dark shadows, with soft wide edges and an island-like pattern. Age cracks form as the whole structure responds to changes in temperature and humidity over decades, and they generate networks that run across the entire surface. Both are normal in a picture of any age, and their absence is more suspicious than their presence.

The systematic evidence is worth knowing because it is unusually well tested. Spike Bucklow, at the Hamilton Kerr Institute in Cambridge, surveyed more than 500 paintings to ask whether crack patterns function as unintentional signatures. Photographs from four art-historical groups, being fourteenth- to fifteenth-century Italian panels, fifteenth- to sixteenth-century Flemish panels, seventeenth-century Dutch canvases and eighteenth-century French canvases, were sorted by thirty people ranging from experienced connoisseurs to college students. Human sorting correlated with the art-historical groupings at roughly 75 per cent; the patterns confirmed or refuted a proposed category at better than 90 per cent; and pairwise discrimination between two categories reached 97 per cent. Automated analysis using Bezier curve parameters achieved 82 per cent overall and 94 per cent pairwise.

Read that honestly. Crack networks carry real information about support, ground, binder and application, and they support or contradict a broad category. They do not authenticate a picture, they do not date one to a decade, and they are disturbed by lining, transfer, heat and pressure, which most old canvases have experienced. Use a crack pattern to raise a question about a claimed origin, then take the question to the technical literature.

Two observations are available to the eye at six feet. On a panel, cracks commonly relate to the grain direction of the wood; on a canvas they commonly relate to the weave and to the stretcher. And a rectangle of visibly different cracking just inside the edges of a canvas is the stretcher bars declaring themselves, which tells you the picture is on its original or a long-standing stretcher.

Edges tell you what has been removed

When a canvas is stretched on a loom or a strainer and tacked, the tension at each tack pulls the threads into a scalloped curve near the edge. That distortion, called cusping, is locked into the fabric and survives later treatment. Cusping along an edge indicates that the edge is close to original. Its absence along an edge, particularly when the other three edges show it, is the standard indication that the canvas has been trimmed, since the cusped zone was cut away.

The Artemisia Gentileschi self-portrait is the textbook demonstration. Exaggerated cusping was present along the bottom edge and some along the two vertical edges, and there was none along the top. Combined with the composition, in which the tip of the martyr's palm and the central pearl of the saint's crown were both missing, the conclusion was that part of the original image had been trimmed from the top. During treatment a new strip of canvas was added to restore the likely original dimensions.

From the visitor's side you cannot see tacking margins, which sit behind the frame. What you can read is the relationship between the composition and the frame rebate. A hand or a foot amputated by the edge, a crown or a hat pressed hard against the top, an attribute cut through, a sitter with generous space on one side and none on the other: any of these is a reason to go and check the catalogue entry. Painters compose to their format, and a format that fights the composition is usually a later format.

On panels, look for a straight vertical or horizontal line of disturbance marking a join between boards, a split running the full height, and a strip of differently aged paint along one edge, which indicates an addition. All three are visible without special lighting once you know to look for them.

The label is a document, and its least readable part is the most useful

A wall label carries what museum staff call tombstone data: attribution line, title, date, medium and support, dimensions, credit line and accession number. Read it as a document with an author and a date, because that is what it is.

The attribution line is the institution's current position and nothing more, and it is often blunter than the museum's own scholarship. Compare it with the online collection entry and, where one exists, the published collection catalogue, which will usually be more cautious and will give the reasoning. A wall does not have room for a hedge.

The credit line records how the object entered the collection: gift of, bequest of, museum purchase, and frequently the name of the fund. That is a provenance statement, and it is the terminal event in the chain. The name in it is a real person or family who will appear in probate records, sale catalogues and the museum's own acquisition files.

The accession number encodes the year. At the Art Institute of Chicago, the first four numbers of the credit line always indicate the year the work was accessioned. Formats differ between institutions, many using a year, then a lot or acquisition group, then an item, so that 1962.24.3 is the third object of the twenty-fourth acquisition of 1962; but the year is nearly always there. Standing in the gallery, that gives you the closing date of the provenance for nothing. It also gives you the key to the curatorial file, which is what you will quote when you write to the department.

Loan labels are a different kind of statement. 'Lent by' or 'Private collection' is wording negotiated with the lender, and anonymity is a request rather than an absence of knowledge. Note the wording precisely, because it is often the only public trace of a current owner.

Hanging position is an institutional argument

Where a picture hangs is a curatorial claim, and it is made without a footnote. A work hung at centre line, alone on a wall, with clear space around it, is being asserted as important. A work above a door, in a corridor, in a stairwell or in a double-hung block is being ranked, whether or not anybody would defend the ranking in print. Adjacency proposes relationships, of workshop, follower, source or rivalry, that the label may not state and the catalogue may not support.

Glazing is informative. Reflective glass over an oil painting usually signals fragility, a recent treatment or a lender's condition, and non-reflective glass introduces a slight dulling and a faint colour cast that you should account for before judging tone.

The frame is rarely the artist's, and in European collections is frequently nineteenth or twentieth century. A frame alters apparent scale, colour temperature and the sense of depth, and museums increasingly reframe to a period-appropriate model, which is a scholarly argument made in wood and gilding rather than in text. Look at the frame, note whether it is likely to be original, and discount accordingly.