I have before me two sheets of statistics about the dining-out and wine-drinking habits of a reasonably depressed American city without much of a culinary tradition, and I must say these have caused me puzzlement. Let me give you some examples.

Every student in the city who eats out at all does so at some point in what is called a 'quality restaurant' (not a food fuel station such as Burger King or Pizza Hut); 97 per cent of them more than 10 times in the month surveyed. Of all restaurant- goers, women diners outnumber men by 8 per cent, and the 18-34 age group eats out most often.

Married people form exactly half of all restaurant-goers. Twice as many people who own their own homes eat out as those who rent, and people with an income below dollars 25,000 (the poverty level for families is about dollars 18,000) ate out 10 or more times in the given month - just 30 per cent fewer than those with an income at least three times as great. Married people are 25 per cent more likely to drink wine with meals than single people, and women are 6 per cent more likely to do so than men.

What are we to make of this? For do we not all assume that the rich patronise 'quality restaurants' more often than the less rich? Or that the single are less likely to eat at home than the married? Or that the divorced, driven by celebration, sorrow or sociability, would be the more likely tipplers?

The quick answer is that the status of the winer-diner (which the survey examines) may be of less relevance than his tastes in food (which it does not). But a more complex answer suggests that all such surveys should begin with the opposite research: that is, a sample of people who seldom or never eat out. These are about a third of the total. Two-thirds of the total, what is more, don't drink wine, and 'quality restaurant' diners form a mere 5 per cent of all those who do eat out.

To me, judging the food in 'quality restaurants' on the basis of the habits of eaters- out and restaurant drinkers is highly unreliable. Such people are as untypical as food writers, for when they (or we) speak of a 'decline' in standards or an 'improvement', they (and we) are already in a peculiar position of privilege in regard to those who almost never eat or drink outside their homes.

This was brought home to me forcefully this week when a group of distinguished literary minds were publicly discussing the 'decline' of high art. The poet Derek Walcott, making his strokes as languidly and expressively as any West Indian cricketer, pointed out that, in his world, one could not speak

of optimism or pessimism about the future of high culture: to do that, one had to have enjoyed it.

In the same way, it is the healthy who most study mortality statistics and most fear illness; the unhealthy would rather not know. And as Hamish McRae pointed out in this paper recently, it is the best educated of the middle classes who have the poorest perception of risk: in food or in any other domain.

If you ask why that is, I can only answer that statistics tend to mesmerise us only in proportion to the way they are presented. Just as an operation presented to us as having a fatal-risk factor of 11 per cent has a success rate of 89 per cent, so a survey of restaurant behaviour that is based on the 1 per cent of foodies and wineys who eat in 'quality restaurants' is not just statistically insignificant, but fatally ignores the reasons why the other 99 per cent do not eat in them.

If you compare the number of column inches devoted to the needs of that 1 per cent, you will note that the Walcott principle prevails: only those who enjoy such a lifestyle have a stake in its maintenance and propagation. The column inches are the product of those with a direct interest in reinforcing their own (often moral) views.

As in dressing or the arts, the most fashionable (and interesting) are those who do not follow fashion, so in our thinking about food or drink it may be that we have much to learn from those who do not scurry after food chic. If writing about food is to have any educational value, it is to the silent majority that it might profitably be addressed.

Yes, this is in part an auto- critique.