CAN we believe a word of what people tell us in our sex survey? That figure for masturbation, for example, looks on the low side and the oral sex findings are hard to swallow.

Our survey was carried out by an experienced, professional polling organisation and was based on the famous "representative sample" of the population. It is not one of those sex surveys where newspaper and magazine readers are invited to phone in or to cut out a questionnaire. These are bound to show that everyone is up to the most peculiar practices because the sample is self-selecting and therefore likely to attract braggarts and exhibitionists.

But some experienced polling organisations using "representative" samples of the population got the result of the 1992 General Election wrong and they have been trying to explain away their failure ever since.

The most common excuse was that people were lying to pollsters. Three per cent of our survey admitted fibbing and presented us with the " 'All Cretans are liars,' said the Cretan" paradox.

The terrible - and to the pounds 500m-a-year market research industry, costly - truth may well be that even if people were honest, the vast majority of opinion polls would still be flawed. This is because up to 45 per cent of those stopped in the street by pollsters refuse to answer questions and more or less politely tell the market researchers to mind their own business.

These are not the "don't knows", who at least try to answer and are marked down as part of a "representative sample", but the "won't says", who walk on and have nothing to do with the silly business.

The growing numbers of refusniks tend to be conservative - perhaps because conservatives tend to be anti-social and less willing than the rest of us to spare a minute for a stranger. Whatever the reason, pollsters have consistently underestimated Tory support since the 1983 election.

If opinion polls on politics can be off beam, how much more suspect will they be when they try to deal with the messy, anxious, private world of sex?

A lot more would be the best guess. But because sex research can never be tested in a general election, pollsters are free to produce gloriously goofy and contradictory findings.

The 1948 Kinsey report shocked an authoritarian United States by "revealing" that one in 10 men was gay and 30 per cent had had some kind of homosexual experience.

Last year the biggest survey of the sexual habits of the British, by the Wellcome Trust, "showed" that just one per cent of men admitted to being homosexual and only 6per cent had had a homosexual experience. Who was right? Kinsey's methods were suspect, but the Wellcome researchers said that Aids-inspired fear may have led to an understandable reluctance to admit to the odd bit of cottaging on a Saturday night.

Kinsey said that only 14 per cent of women had had multiple orgasms. But this year one Dr Marilyn Fithian at Long Beach University's Centre for Marital and Sexual Studies (I'm not making it up, it exists) "found" that 75 per cent of women from her self-selecting - and therefore meaningless - sample had had multiple orgasms. One volunteer in her lab had 134 in 50 minutes without needing a tea break.

The same pattern appears whatever surveys one looks at. Searching through the findings of sex polls is like trying to find consistency in an election manifesto.

Example: A 1993 study by Drs Samuel and Cynthia Janus (and no, I'm not inventing their names either) found that 60 per cent of people said that sex was better after marriage and 83 per cent said their marriages were either great or good. But in the same year Cosmopolitan reported that half of married men cheat on their partners, and in 1994 a survey for New Woman said 75 per cent of women no longer believed in marriage.

The confusion may stem from people not understanding the pollsters' questions. How many of our sample knew what the words anal sex or masturbation meant, even if they had performed them in practice frequently?

A pilot project for the Wellcome study asked one 30-year-old what he thought about heterosexuality. "Well it's all the same to me," came the unreconstructed reply. "Heterosexual, bisexual, they're all bloody queers."

Many argue that sex surveys are worthwhile because they give comfort to people who would otherwise suffer in silence. Alfred Kinsey's findings may not have been true but they brought relief to many isolated and persecuted homosexuals. But supporting spurious research because it may produce desirable social consequences is a dangerous game.

For years gay activists bolstered their cause by chanting that Kinsey had proved that one in 10 men was homosexual. When the 1994 Wellcome findings on homosexuality were released, they were seized on by grateful Conservative MPs and certain newspapers and used in their successful campaign to stop the homosexual age of consent being lowered to 16. Homosexuality was a rare perversion, after all, the campaigners argued, and gay people had exaggerated their numbers for political reasons. So the argument became about numbers not rights, as if private life should be regulated by plebiscite.

The cynical view is that newspapers will find what they want to find in opinion polls. But the problem with the journalists who promote polls is not that they are mercenary manipulators - although polls are certainly used to manipulate "the agenda" - but that they have a desperate need to appear objective.

Journalists need "facts" to write about, even if the "facts" turn out to be spurious. Hard numbers are reassuring and far easier to collect than the conflicting and fragmentary reports from an uncertain world, which, nevertheless, contain the only "real truth" we can find.

"It is the spirit of the age," sighed Gore Vidal, "to believe any fact, however suspect, is superior to any imaginative exercise, no matter how true." Opinion poll findings are the most suspect of facts; and sex can be, among other things, an imaginative exercise.