WHEN VENUS rose up from the ocean on her sea shell and the sky rained roses and Botticelli painted her to show all later men how fine a day can be, he did not paint the roses quite as carefully as I would have liked.

I am almost sure they are also roses, but possibly they are centitolias? You would think the painter would have paid more attention since the roses are, needless to say, the most remarkable thing in the picture. There are plenty of Renaissance Venuses floating 'about, and the Lord's own plenty of sea shells, but I do not think there are many such roses in pictures so old.

And yet if you ask an art critic about the roses, or try to find out more about them from all the words that have been written on this painting over the centuries, you will discover that (for some unfathomable reason) most of the stuff is about Venus or some other minor feature of the work such as color, sex, volumes, etc.

Nothing about the roses.

All this was on my mind when I visited my friend the anthropologist (Dr. E. Richard Sorenson) this week. We have been on this very subject before -- the failure to notice the roses and to get off on minor things like the naked goddess -- and he said it is a real problem for anthropologists and, needless to say, for Science itself.

It is not my fault, certainly, that so many students of the painting of Venus have got off in little byways and ignored the big question of the identify of those blooms.

Sorenson said -- and for some reason he is not so interested in the Botticelli roses as I am -- the identical problem arises constantly in anthropology. Sorenson, 48, is the director of the National Anthropological Film Center here, and he has an almost panic passion for filming the few remaining isolated cultures left in the world.

Cannibalism is going out. How hard it is, nowadays, to find a society without government, money, commerce, cloth, metal, crime, disabled fire department personnel, etc.

And yet Sorenson has lived for months on end among people who know nothing of any of these things, and he has hundreds of thousands of feet of film to prove it.

His operation is complex, but we did ruminate, once again, on the agony of finding incomplete records on some topic of tremendous interest (like the Botticelli roses or, in Sorenson's case, the way the toddlers of the New Guinea highlands wield axes and knives without ever hurting anybody or themselves).

Sorenson has given the question far more attention than most, and I never visit with him without thinking how central the problems of anthropology are to almost everything else in the world. There are three general ways to film a society you know nothing about to begin with.

You can show up with your camera (you should of course become friendly with your subjects first) and shoot whatever seems interesting.

Almost any anthropologist would shoot footage of a tribal dance or cooking the priest in the pot. (In actual fact, that is not fair, for many primitive societies are enormously sophisticated in personal relationships and many are more agreeable to strangers than our own society is.)

A second general approach is to decide, right here in Washington, that you will study breast-feeding among some isolated tribe, say, in order to compare it with customs elsewhere. The breast-feeding of babies is, needless to say, a matter of tremendous complexity and significance. You do not, of course, barge in like an idiot, but after much prayerful thought you may decide on a film about feeding.

You shoot that and come home.

There is nothing wrong about either approach, except that each is limited. Are they not like the Botticelli painting of Venus, which shows the goddess and brushes off the roses that did not greatly interest the painter?

A later student of the same primitive society (and it may have ceased to exist, as so many have, for in the past 10 years more have disappeared than now remain) may not care at all about either the pot or the breast-feeding. He may be vitally interested in the tribe's transportation patterns or how they prepare daily gruel or how they thatch a hut or how they treat 13-year-olds.

How grievous a thing it is for such a researcher to find earlier films that show (like the Botticelli roses in the corner of the painting) that just beyond the great breast-feeding focus of the film a woman is thatching a roof. And then (as the woman shifts to feed her baby) the thatching operation is out of range and is never seen again.

Of course, as Sorenson says, you can't have a film record of every moment and every place throughout eternity. But you can be aware that your record may be the only one in existence, when the society has completely vanished.

And this brings us to the third general approach:

You can shoot first and ask questions later.

An anthropologist cannot know, in meeting a society totally different from his own for the first time, what is important and what is not.

Sorenson had no way of knowing (to give a simple example) that a facial expression of serene sadness may be, in fact, the customary expression of anger.

By filming a great deal that may seem at first not very interesting, you may discover something you had not noticed. In one society babies are not separated from a mother's body their first three years -- she is constantly available.

If you just dropped in and shot what seemed interesting in five days, you might never notice this astonishing pattern, that even a mother preparing food, even a mother carrying a heavy lead of wood, keeps her baby against her body all the time.

Thus you get all the pictures you can, avoiding rash judgments that this or that is not interesting, and you put the films in a safe place where they can be studied -- frame by frame -- for years to come.

A written account is all very well, but any researcher (or any reader) has to say to himself:

Do I have confidence in this writer? Is he telling it the way "it really was" or is he not?

Even the most objective anthropological film is also limited -- few film-makers will shoot the daily rising of the sun, though that is certainly part of the truth -- but at least it is not filtered through the reporter's own cultural bias and vocabulary and set of assumptions.

I think the world of Sorenson's brains and modesty and constant awareness that all the truth he is ever going to find is going to be a partial truth.

In art the focus is everything, but in historical records the undifferentiated fullness is all. Time after time the stone that the builder rejected turns into the capstone.