LONDON, ENGLAND.- Few pictures have been the focus of as much attention as Leonardo da Vinci's two versions of The Virgin of the Rocks - owned by the National Gallery, London and the Musée du Louvre, Paris. Generations of art historians have wrestled with conflicting explanations of these pictures' histories and their mutual relationship. Now a new discovery by the National Gallery is going to re-ignite the debate.

Curators, conservators and scientists have used infrared reflectography to discover two distinct underdrawings beneath the surface of the National Gallery's 'Virgin of the Rocks', sometimes dismissed as merely a copy of the Louvre's version.

The first of these underdrawings does not correspond at all to the image that we know so well today, and x-radiography shows that none of Leonardo's first design was ever painted. The second is for 'The Virgin of the Rocks' as it was finally executed, but even in this there is evidence of several considerable changes of mind. This shows Leonardo thought the composition through anew for his second version - no-one can ever call the London painting a straightforward copy again!

The 'Virgin of the Rocks' was commissioned in 1483 by the Milanese Confraternity of the Immaculate Conception, and was intended to form part of their oratory altarpiece. However it was twenty-five years until a painting of this subject was finally placed in the chapel. In the interim, Leonardo had painted two versions of the composition: the first (in the Louvre) was probably sold in the 1490s to a private client after a financial wrangle with the Confraternity; and a replacement - the painting that now hangs in the National Gallery - that was installed in 1508.

The first underdrawing in the London painting shows a kneeling figure. Her downcast gaze and pious gestures, one hand held modestly to her breast, the other dramatically outstretched with the fingers meeting the picture edge, indicates that when Leonardo was first asked to paint a second picture he decided on a new composition: an Adoration of the Christ Child. The baby Jesus cannot be detected, but this could be because Leonardo abandoned this idea before he came to include the Child.

Further evidence for this theory comes in the form of a miniscule Leonardo drawing at Windsor - a compositional idea for an Adoration of the Child with an arch-topped panel, exactly like the support of the National Gallery's 'Virgin of the Rocks'. Until now this little sketch has been unexplained in relation to any known commission. Fascinatingly, the design Leonardo abandoned relates closely to several of his paintings of the 1490s. Most intriguingly, he re-used his drawing for the head of the Virgin to paint the disciple, Saint Philip, in 'The Last Supper'.

So why did Leonardo abandon his first underdrawing to revert to 'The Virgin of the Rocks' as he had already painted it? Perhaps he simply found his new idea unsatisfactory, or it may have been a simple matter of time; with so many other commissions pending, was it just easier to copy an earlier work? Or perhaps the members of the Confraternity insisted upon a copy of his first 'Virgin of the Rocks', rather than an original composition? We will probably never know the answer, but the question will perplex art lovers and historians for many years to come.