Journal · Attribution · November 2026

A signature is the weakest evidence there is

It is the first thing most people look for and the last thing a specialist relies on. A signature is a few grams of paint on the surface of an object that may be two hundred years old; it is the easiest part of a painting to add, and the part most often added by someone other than the artist. Attribution is built out of other material entirely.

Why it carries so little weight

Three reasons, and they compound. First, many genuine works are unsigned — signing a canvas only became general practice in the nineteenth century, and plenty of artists never did it consistently. Second, many signed works carry a signature added later by a dealer or an owner who was sure of the attribution and wanted the market to agree. Third, a forger who can imitate a style can certainly imitate a name, and the name is the part that takes five minutes.

So a signature proves almost nothing on its own. What it does do is raise a question: when was this applied, and by whom? That question is answerable — under magnification, a signature sitting on top of a varnish layer was added after the work was varnished, and that is a fact about the object rather than an opinion about it.

The signature is a claim. Everything else on the object is evidence about whether the claim is true.

What attribution is actually built from

• The hand. How the paint is laid down — the handling of edges, the construction of a passage, the habits an artist repeats without thinking. This is what connoisseurship means, and it is comparative: it requires having looked closely at many secure works by the same artist.
• The support. Canvas weave, panel species and construction, ground layer, colourman’s stamp. These date the material and can rule an attribution out absolutely, which is more than style can usually do.
• The documents. Ownership history, dealer records, exhibition catalogues, old photographs, correspondence. A work that can be traced back to the artist’s studio or lifetime is in a different category from one that cannot.
• The literature. Inclusion in the catalogue raisonné, or acceptance by the artist’s foundation or the recognised scholar. For many artists this, and only this, is what the market will accept.
• The laboratory. Pigment analysis, dendrochronology, infrared and X-ray imaging. Expensive, and usually a last step — but definitive in one direction: a pigment invented in 1935 settles the matter for a painting said to be from 1890.

Reading the catalogue language

Auction catalogues use a vocabulary that looks like praise and is in fact a precise scale of doubt. In most Western houses it runs roughly like this, from strongest to weakest.

• Artist’s full nameIn the house’s opinion, a work by the artist.
• Attributed toProbably by the artist, in whole or in part, but with less certainty.
• Studio of / Workshop ofMade in the artist’s workshop, possibly under supervision, not necessarily by their hand.
• Circle ofBy an unidentified hand close to the artist, of the period, showing their influence.
• Follower ofIn the artist’s style, not necessarily a pupil, possibly later.
• Manner ofIn the artist’s style but of a later date.
• AfterA copy of a known work by the artist, of any date.

The exact wording differs between houses and each publishes its own glossary, which is worth reading before you bid. The price difference between the first line and the third is routinely a factor of ten or more, and the visual difference can be very small.

What this means if you own something

If a work in your family has always been “the Corot” or “the Aivazovsky”, the useful question is not whether the signature looks right. It is what else exists: where it was bought, what came with it, what is written on the back, whether anyone has ever written it down. That material either supports the name or it does not, and finding out costs far less than discovering the answer in the middle of a sale.

And if the answer turns out to be “circle of” rather than the name on the frame, that is still a real object with a real value, documented honestly — which is a better position than an optimistic attribution nobody will underwrite.