The only problem is that this definition also leaves out some very good installations. Consider Richard Wilson's 20:50. It consists of a lake of sump oil that uncannily reflects the ceiling of the gallery. Spectators penetrate this lake by walking along an enclosed jetty whose waist-high walls hold the oil at bay. This 1987 work was originally set up in Matt's Gallery in east London, through whose windows one could see a bleak post-industrial landscape while standing on the jetty. The installation, awash in old engine oil, could thus be taken as a comment on Thatcherite destruction of manufacturing industries. Then something very interesting happened. Thatcher's ad man Charles Saatchi put 20:50 in his windowless gallery in west London, depriving it of its context. But the Thames and Hudson definition does not allow that this 20:50 is an installation because it wasn't created for that space. This is silly: it would be better to say there were two installations - the one at Matt's and the other at the Saatchi Gallery.

Or think about Damien Hirst's In and Out of Love. In this 1991 installation, butterfly cocoons were attached to large white canvases. Heat from radiators below the cocoons encouraged them to hatch and flourish briefly. In a separate room, butterflies were embalmed on brightly coloured canvases, their wings weighed down by paint. The spectator needed to move around to appreciate the full impact of the work. Unlike looking at paintings or sculptures, you often need to move through or around installations.

What these two examples suggest to me is that we are barking up the wrong tree by trying to define installations. Installations do not all share a set of essential characteristics. Some will demand audience participation, some will be site-specific, some conceptual gags involving only a light bulb.

Installations, then, are a big, confusing family. Which brings us to the second question. Why are there so many of them around at the moment? There have been installations since Marcel Duchamp put a urinal in a New York gallery in 1917 and called it art. This was the most resonant gesture in 20th century art, discrediting notions of taste, skill and craftsmanship, and suggesting that everyone could be an artist. Futurists, Dadaists and surrealists all made installations. In the 1960s, conceptualists, minimalists and quite possibly maximalists did too. Why so many installations now? After all, two of this year's four Turner prize candidates are installation artists.

American critic Hal Foster thinks he knows why installations are everywhere in modern art. He reckons that the key transformation in Western art since the 1960s has been a shift from what he calls a "vertical" conception to a "horizontal" one. Before then, painters were interested in painting, exploring their medium to its limits. They were vertical. Artists are now less interested in pushing a form as far as it will go, and more in using their work as a terrain on which to evoke feelings or provoke reactions.

"Many artists and critics treat conditions like desire or disease as sites for art," writes Foster. True, photography, painting or sculpture can do the same, but installations have proved most fruitful - perhaps because with installations the formalist weight of the past doesn't bear down so heavily and the artist can more easily explore what concerns them.