Did the Old Masters have their heads in the clouds?

'On the whole, Thomas Gainsborough hadn't the faintest idea what he was looking at'

Do artists paint what they see, or what they want to see? Is painting what it so often seems to be, an innocent representation of the world? Interesting questions, raised again by an exhibition in Edinburgh of Constable's paintings of clouds. Constable was a keen amateur meteorologist; in scientific, as well as in poetic ways, he was deeply interested in what skies and clouds can tell us. He had the fortune, too, to live at a time when the modern classification of cloud types had just been codified, and incorporated this model into his painting in fascinating ways.

Do artists paint what they see, or what they want to see? Is painting what it so often seems to be, an innocent representation of the world? Interesting questions, raised again by an exhibition in Edinburgh of Constable's paintings of clouds. Constable was a keen amateur meteorologist; in scientific, as well as in poetic ways, he was deeply interested in what skies and clouds can tell us. He had the fortune, too, to live at a time when the modern classification of cloud types had just been codified, and incorporated this model into his painting in fascinating ways.

This information has come to light in a really interesting study by the meteorologist John E Thornes. Starting from the fact that Constable often describes the exact weather conditions when he was painting a landscape in the open air - one characteristic description runs "Sepr.21 1822. Looking South brisk. Wind at East, Warm and fresh. 3 o'clock afternoon" - Dr Thornes has examined the plausibility of Constable's observation of the weather. It is strikingly accurate, on the whole; in many of his landscapes, it is possible to say what weather had preceded and followed the moment of observation.

All very fascinating. One newspaper had the bright idea of asking a weatherman to go round the National Gallery and let us know whether artists before Constable got it "right or wrong" when they looked at the sky. He agreed with Dr Thornes that, on the whole, the Old Masters hadn't the faintest idea what they were looking at. Poor Gainsborough, whose skies were considered famously beautiful, came in for particular stick. The clouds in Mr and Mrs Andrews look more like the smoke from a bonfire than any meteorological phenomenon. As for The Watering Place, that formerly sublime pastoral vision, the weatherman complained that he couldn't tell whether the artist had intended to paint a sunset or a sunrise.

At which point, I think we might pause for a little thought. It's certainly true that Gainsborough's skies don't look like any sky one has ever seen, and you couldn't tell what the weather was like from looking at one of his paintings; it would be like asking an ornithologist to identify the birdsongs in Messiaen. But then, Gainsborough's trees don't look very much like trees either - not surprisingly, since he wasn't looking at trees while he was painting them, but at little models of cork and moss, rigged up in the studio. Nor do his people look very much like people; gorgeously elongated, pale and perfected, his sitters look as if they know about anatomy but prefer not to indulge in anything so vulgar. In short, Gainsborough just wasn't looking.

It's certainly true that art has often found a new and more perfectly representative way of looking at the world a great stimulus to invention and ingenuity. The understanding of perspective, for instance; the obsession with the accurate representation of a particular moment of light, which came with the 19th-century craze for plein air painting; even the photographic styles of Sixties pop art; all these fed back into the art and added a new excitement to creativity.

But it's wrong to think that art necessarily improves in any way, or that a truthful painting of a sky full of clouds is any better than an artistic convention. If you look at a lot of the very best 18th-century painting, you might point out, soberly, that the clouds do look a lot like cotton wool. But then you would have to agree that babies don't float on clouds; that women are rarely eight feet tall, with skin indistinguishable from the marble on which they so comfortably recline; that landscape is never so impossibly green, smooth and airy.

The truth is that, along with the incentive to represent the world, there has always been a desire to paint the world inside the head. There's no point in comparing Turner's wildest fantasies with Constable's sober, scientific, handsome renderings. They are trying to do something different. One has his eyes wide open; the other, to see something different, has closed them tight. We can certainly praise the excellence of Constablewithout thinking for a moment that one sort of painting is necessarily better than the other. Or even, in a real sense, any more accurate.

Does it matter? I wonder. There is a childish sort of objection to lots of art, from Parmigianino's madonnas to Braque's still lives, that "it's just not realistic"; that people, on the whole, don't sing coloratura arias for 10 minutes just before they die of consumption; that pigs are not going to mount a revolution and take over a farm. In one sense, these things are not true and will never be true. But anyone who understands painting will see that this is not the same as formal logic. There are different sorts of truth.

hensherp@dircon.co.uk