Arts: On the one hand, but on the other . . .: Juliet Wilson-Bareau selected 87 paintings by Goya for the Royal Academy. That she rejected many more may alarm a number of museums and private collectors. Dalya Alberge reports

Juliet Wilson-Bareau, curator of the Goya show at the RA, inspecting the sketch for Autumn, 1786, whose attribution had long been doubted; some said it was a fake - partly because it had not been sold with related works in the 1890s. But, convinced it was an original, Wilson-Bareau tracked the painting down to an American collection in Williamstown, Massachusetts and proved, through a mass of graphite underdrawing, that this was indeed a Goya. It is one of several Goyas whose attributions she has confirmed. 'Where there is vigorous underdrawing underneath,' she says, 'you can be sure there was a vigorous painter at the back of it' Photograph: Nicholas Turpin

Goya was a god among his fellow artists. In the 19th century, lesser mortals made so many paintings in his image that the hundreds of imitations, copies and fakes that appeared began to be confused with the real thing. And in the confusion, some genuine Goyas were somehow de-attributed.

Although warning bells were sounded about the number of copies as early as the 1880s, and many works have since been 'rethought', would-be Goyas remain in collections worldwide. The Metropolitan Museum in New York is coming to terms with the fact that one of its most prized paintings, Majas on a Balcony, is probably by Somebody Else.

Goya attribution is a minefield, as Juliet Wilson-Bareau, a leading Goya scholar, was all too aware in curating a major show on him, the first in London for nearly 30 years. Some 87 small-scale images - no less incisive and powerful than his large works - go on display in London, at the Royal Academy on 17 March, after a run at the Prado in Madrid which proved so popular, it was extended for a further fortnight.

Taking the view that, of all the paintings that have ever been labelled 'Goya', as many as half could be discounted, she submitted all her chosen paintings to rigorous censorship. Every brush stroke, every nuance of light, every detail of expression was examined; she sought in particular vital signs of life in the figures that people the master's vicious satirical attacks on Society and the Church. Would-be Goyas sought to reproduce the savage emotion of his scenes of war and famine, of the madhouse and the bullring, but the images pale into insignificance beside the originals: their satire is flat, their characters are devoid of life. So often, particularly with 19th-century copies, there's a sense of ham acting in the grimacing - 'too much rolling of eyes', as she puts it.

While she de-attributed dozens of works, Wilson-Bareau was able to establish with absolute certainty the authenticity of a number of pictures which scholars had long doubted. Among them was his sketch for Girls with Water-jars, 1791, which had been seriously doubted because details in the sketch are missing from the final tapestry cartoon image. But when Bareau-Wilson had the final composition X- rayed, she found those details: the X-ray showed that an old woman, who is in the sketch but missing from the cartoon, is in the cartoon's underpainting. The artist had changed his mind about including her in the composition.

Particularly exciting was authenticating a Goya in a private collection, Hannibal the Conqueror, his first documented work. Initially, she explains, 'I couldn't believe that this was by Goya . . . It is an academic classical subject, which was not something he went in for.' But there was something about it that encouraged her to persevere and to persuade the owner to send it to the Prado, where it was cleaned and the overpainting removed. Suddenly, Goya's vigorous brushwork emerged. Heavy-handed restoration had concealed the details: a tiny hand on Hannibal's shoulder is painted so intricately that every finger is indicated. 'It's perfect,' she says, explaining that in many Goyas the slightest, flicking strokes create hands that really grip or point.

Its attribution was further confirmed by the recently discovered notebook, 1770-85, filled with sketches for Hannibal and other works: Wilson-Bareau not only tracked it down but has been instrumental in the Prado acquiring it. Due to her decision to have the Hannibal sketch cleaned at the Prado, the painting to which it relates was also located: a curator, who had happened to spot it in the lab, recognised the same image in a collection in Spain. That was attributed to an unknown 18th-century artist: it turned out to be the actual picture that Goya submitted for a competition in Palma.

Of a further 60 or so works which Wilson-Bareau considered for the show, many were rejected because she found herself unable to dispel even the slightest misgivings. In some cases, doubts arose because there was another version somewhere.

That was one of the most damning pieces of evidence against the Metropolitan's Majas: the Met decided to 'out' the picture as a powerful version in a private Swiss collection has a provenance that can be traced back to an 1812 inventory. With two Goyas of children at play at Pollok House, Glasgow, there are up to five versions: she believes that each of them is by a different hand.

It is one thing for Wilson-Bareau to test Goyas, quite another to tell owners that their painting had failed it. As she puts it, 'I can't go around telling people their pictures are not by Goya. Even with museums, it's a delicate subject.' With Goya worth millions, both fortunes and reputations are at stake. Sometimes, she hinted at unease over a work; occasionally, she was blunter.

Until a catalogue raisonne is produced, however, owners of suspect Goyas may console themselves that Wilson-Bareau may be wrong. Not all scholars accept the question marks over the Majas on the Balcony or the Glasgow Goyas, and Wilson- Bareau is the first to admit that 'no one is infallible'. She cites the example of two paintings in the Ashmolean, Oxford. Like many non-Goyas, they were attributed to Eugenio Lucas, who was 'doing' Goyas from the 1830s. In the 1950s, they were reattributed as Goyas, yet in the 1990s, Wilson-Bareau has rejected them. 'There's no way they're by Goya,' she says.

But then that's the nature of art scholarship: it's not an exact science. You win some; you lose some.

Goya: Truth and Fantasy, The Small Paintings, Royal Academy of Arts, Piccadilly, London, W1 (071-439 7438): 17 March-12 June.

(Photograph omitted)