Comparison of individual genes showed that the relationships between plant families, which had hitherto been widely assumed, were in some cases wildly inaccurate; more than that, the most surprising kinships existed between the most unlikely flowers and trees.
Gazing on the red blooms of the Asian lotus as they emerge from the still waters of a pool, for example, you are not immediately put in mind of Berkeley Square. But you might well be, Kew's molecular biologists have discovered.
The huge flowers of Nelumbo nucifera, for thousands of years a sacred plant in India, China and Tibet, are not related to the waterlily family which they resemble, and with which botanists have long and naturally associated them. Their closest relative, it turns out, is Platanus hybrida - the familiar and hardy plane tree of London's squares, so common because until the Clean Air Act got rid of London's smog more than 40 years ago, they were the only things that would grow there. The sacred water flower of the East and the giant tree of Britain's capital are first cousins.
In revealing many more such unusual relationships, the exploding science of genetics has made possible a complete - and now accurate - reclassification of all the families of the world's flowering plants and trees, which the scientists at Kew have led and which has just been completed. It represents for the first time an evolutionary tree of plants which is certain to be accurate.
It has been based on comparing their DNA, their genetic code, rather than their morphology - their appearance and physical characteristics, which is all that plant taxonomists have hitherto had available to them.
People have been classifying plants since prehistorical times, of course: which ones provided food, which were poisonous, which were medicinal.
Medieval herbalists made bigger classifications, but it was not until the Age of Enlightenment at the end of the 18th century that a complete and systematic classification of plant families was first attempted, by the Swedish Naturalist Carl von Linne, known as Linnaeus.
His classification was considerably improved by 19th- century French botanists, such as Decandolle and Jussieu, who established the families with which we are now familiar, such as the compositae, the daisy family, the brassicae, the cabbages or mustards, the rosaceae, the rose family, and the orchidaceae, the orchids.
Twentieth century botanists have continually refined these classifications. The trouble has been, says Dr Mark Chase, the man who has led the Kew DNA team, that no two of them have been exactly the same.
"There have been three major classifications of the world's plants in the last 20 years and each of them has had different ideas, not only between each other, but between different versions of the same classification," he said.
For example, in 1981, the American Arthur Cronquist divided all plants into 321 families; the next year, his fellow American Robert Thorne divided them into 440 families; and in 1997, the Russian Armen Takhtajan divided all plants into 598 families.
"These men were trying to assess the plants' genetic information from what they could see with the naked eye," Dr Chase said. But he and his fellow scientists from Kew and from around the world have now done something new: they have gone to the genetic information directly.
New techniques of molecular biology that became available in the past decade have allowed them to sequence, or identify, individual genes in plants' DNA and then assess them for similarities across species.
They have done this with three genes for each of 565 plants representing all the world's flowering plant families, the first and most important being the gene responsible for rubisco, the enzyme that controls photosynthesis, the essential process by which plants convert sunlight into energy.
The relationships between plant families indicated by the rubisco gene has been exactly repeated with the next two genes, convincing Dr Chase and his colleagues that at last, 200 years after Linnaeus, they have the true picture of how all the world's plants are related.
They now class them in 464 families. Botanists will have to rewrite their floras as a result. Not only is the lotus related to the plane tree rather than the water lily. Roses are not related to saxifrages or the bean family, as had been thought, but to the buckthorns, the nettles and the figs. Orchids are not related to lillies, as once thought, but to the yellow star-grasses.
So from next month, when the new classification is published, a rose will still be a rose; but much else in the plant kingdom will be very different.
IT WAS an 18th-century Swedish botanist who first developed a scientific system for classifying all plants, and subsequently all living things.
Linnaeus was born in 1707 as Carl von Linne, the son of a country curate, and was obsessed with flowers from the age of eight. As a research student at the University of Uppsala in 1730 he realised that the scientific arrangement of plant species was inadequate, and began to sketch out his own.
Linnaeus invented binomial nomenclature - the system of using two Latin names to describe plants and animals, the first describing the genus, or related group, and the second the species itself.
Thus herb robert, the common pink flower of English woodlands, is (Geranium robertianu : it is a member of the Geranium genus, like many others, but robertyianum describes this particular one. This naming system is now in universal use for all living organisms.
The criterion of classification Linnaeus used for plants has been superseded. He classified them according to their sexual parts, such at the stamens or anthers (the male parts), and how many each had: a plant with five anthers would be in the pentandriae. The modern division of the plant kingdom into large related families, such as the cabbages (brassicae) or the daisy family (compositae) is a product of 19th-century botanists.
The new classification supersedes both: it divides the plant world into families, but for the first time does it with precision.