Sometimes trusting what scientists tell us can be a bit difficult. One day we are told that artificial sweeteners help prevent obesity; the next, that they cause it.

The generous explanation for such seesaws is that science is always developing our understanding. But there is a more sinister concern: fraud.

No fewer than 15 per cent of scientists at the National Institutes of Health, the US Government’s top health laboratory, recently admitted to bending data to fit their theories.

The myth is that science is the noble search for truth. The reality is that scientists are selfish. In the old days scientists often published secretly to safeguard – and profit from – their discoveries. On writing a paper, a researcher at a university might deposit it in a college’s safe, publishing it only if someone else made the same discovery later. The first scientist would release his data to make sure everyone knew he had got there first.

The same principle was behind using codes to protect intellectual property. In 1676 Robert Hooke published his law of elasticity as a Latin anagram: "ceiiinosssttuv". This made sure that he would be credited for the idea, which he later revealed to be "Ut tensio, sic vis": stress is proportional to strain.

Inevitably this secrecy caused problems so during the 17th century Robert Boyle created a club within which scientists did reveal everything to their fellows. People still worried about being scooped but, as members kept their findings secret from non-members, the insiders enjoyed huge advantages. The name of this association was the Royal Society.

The conventional narrative holds that, as the advantages of pooling knowledge became obvious, all scientists adopted the Royal Society’s conventions: now scientific papers are published freely. But that is not quite true. In fact, scientific journals are as closed as the Royal Society once was. The gatekeeper is "peer review". Experts judge if the experiments the manuscripts describe are credible.

But how, without having witnessed the experiments, can experts determine that? Reviewers have to trust the authors to have told the truth. Consequently, the most important part of a paper is the name at the top. If a well-known scientist submits a paper, it will probably be accepted; if an unknown submits one, it will probably be rejected. Science is still a closed club – partly to ensure that only accurate papers are published, but largely to prevent fraud.

But peer review carries dangers. First, it allows dunderheads to block unexpected ideas. Everybody in science knows of researchers such as Barbara McClintock, who won a Nobel prize in 1983 for discovering gene-jumping, a process by which scraps of DNA move about the genome. She had to publish her findings informally, in the annual reports of the Carnegie Institution, because she could not persuade peer reviewers to accept them. Moreover, peer review is slow, and allows unscrupulous reviewers to plunder their competitors’ papers and to block their publication.

As we enter the Wiki-world, peer review will lighten. People once paid for hard copies of journals but now free periodicals such as PLoS Biology, published by the Public Library of Science, proliferate online. They are still peer-reviewed but soon reputable scientists will start to publish their own electronic papers. The convenience will be irresistible.

Some form of peer review will need to survive, to deter fraudsters, but it will probably resemble the one practised by the Proceedings Of The National Academy Of Sciences Of The United States, in which, essentially, distinguished friends vouch for each other.

And a good thing, too. Peer review was always an illusion, providing a deceptive imprimatur of objective truth. Less formal arrangements will remind us that new science is always provisional – and that validation comes only after publication, when others try to reproduce the work.

Terence Kealey is vice-chancellor of the University of Buckingham.