Seeing, Not Predictions

In 1949 an old man in Switzerland was asked to introduce a book of Chinese divination to the English-speaking world, and before he wrote a word he asked the book what it thought of the idea.

Jung was in his eighth decade, which is how he put it himself. The book was the Yijing, the Book of Changes, in Richard Wilhelm's German and about to appear in English. He took three coins, threw them six times, and got the fiftieth hexagram, Ting, the Cauldron — with two moving lines, which turned it into the thirty-fifth, Chin, Progress. Then he did the thing that has irritated people ever since. He read the answer as the book talking about itself: a cauldron is a ritual vessel holding cooked food, the food is spiritual nourishment, and the book was telling him what it was. Progress, he decided, was where it was headed, which in 1949 meant the American book market.

Partway through writing he put a second question, about his own conduct in writing at all, and got the twenty-ninth hexagram, K'an, the Abysmal — a pit — changing into the forty-eighth, Ching, the Well. He notes the run himself and leaves it there: vessel, pit, well. Then he published all of it under his own name, at the height of his reputation, in another man's book.

It is easy to be embarrassed on his behalf, and better to read what he wrote around it. He concedes that he is no sinologist and that this can only be a testimonial of his own experience; he records that Chinese scholars had told him to his face the book was a collection of obsolete magic spells; and near the end he states the case against his own procedure more crisply than his critics have since — anyone clever, he writes, can turn the whole thing around and show that he projected his own contents into the hexagrams.

He used the thing, published the result, and named the strongest objection to himself. Whatever that is, it is not credulity.

It is not honesty either, and that must be said plainly. Jung is a poor witness about his own conduct across three decades. His 1902 dissertation was built on séances with his own cousin, whose relation to him he did not disclose and whose age and dates he adjusted. In 1909 he wrote to a former patient's mother denying he had ever been more than her doctor, confessed the lie to Freud as a piece of knavery, and in the same letter asked Freud to write to her vouching for his perfect honesty. In 1949, asked about an essay contrasting the Aryan and Jewish unconscious to the latter's disadvantage, he said the passages had been tampered with. Some circulating versions had been: a 1945 translation and a wire story put words in his mouth. But the passages that damn him are his own, unaltered, in his own authorised collected works, and he never publicly retracted them. The book most people read as his honest account of himself is one he wrote should appear under his collaborator's name and not his own.

A man who insisted he was an empiricist observing phenomena and refusing metaphysics, unreliable about the one phenomenon he could not observe from outside. The essay needs only the first half: a compromised witness describing a procedure and its outcome is still describing a procedure and its outcome. Nothing below rests on his character, and whatever an instrument does for the person holding it, it does not make them good.

One caution about the book he was holding. The standard English Yijing was translated at Jung's urging by a member of his circle who did not read Chinese, from a German version sinologists have faulted for a century, and published with his foreword by the foundation his patrons built. Nothing below leans on it.

Which matters, because nearly every argument about divination since has missed what Jung was doing — each conducted on a premise both sides share and neither examines.

The premise is that the question is about the future.

One camp says these systems tell you what is coming. The other camp says they demonstrably do not, and produces the arithmetic. The second camp is correct. There is no evidence that yarrow stalks, coins, cards, entrails or planetary positions carry information about future events, and there is a great deal of evidence about how easily people convince themselves otherwise.

And here is where an essay like this usually cheats. The tempting move is to say the traditions never meant it that way — that the debunkers are demolishing a straw man. That is false, and anyone who knows the material knows it is false.

Divination in China was mainstream for more than two thousand years — not a folk survival at the edge of respectable life but central, elite, and state-sanctioned. Richard Smith, whose history of the Yijing is the standard account of what the book was used for, is blunt about it: for most Chinese intellectuals, consulting the Changes was the most useful tool available for understanding the world and their place in it, and the educated complaint was never against divining but against venal practitioners. Zhu Xi, the most influential commentator the text ever had, held that it was primarily a work of divination.

Nor were people credulous. Mambila diviners calibrate a spider with questions they already know the answer to, and discard the ones that fail. Among the Nuosu, Ze Hong and Joseph Henrich found practitioners weighing accuracy and voicing scepticism. Three thousand years earlier the Shang were carving outcomes into the bone beside the forecast.

So the prediction business was real, it was everywhere, and it was audited.

What is also true, and what the modern argument never touches, is that the same traditions kept distinguishing two ways of using one instrument. And it was insiders who drew the line.

In 1973 a tomb was opened at Mawangdui in Hunan, sealed since the second century before the common era. Among the silk manuscripts was a text nobody knew existed. It is called Yao, the Essentials, and part of it is an argument between Confucius and his student Zi Gong.

Zi Gong puts the objection anyone would put. You taught us that where virtue fails people turn to spirits, and where planning falls short they turn to divination. So why, now that you are old, do you delight in this book.

The reply is the most useful sentence in the controversy. Confucius says he is on the same road as the scribes and the spirit-mediums and ends up somewhere else. Elsewhere he says he gets seventy of a hundred prognostications right — his own, not the profession's — and is untroubled by it, because accuracy is not what he is there for.

He does not abandon the practice. He ranks it: prayers and divinations, he says, come last.

The manuscript lay in the ground for two thousand years, so it cannot have been retrofitted by later apologetics. Whatever the historical Confucius said, this is what the tradition took its founder's position to be.

Xunzi draws the same line two centuries later, harder and colder. His gentleman still divines — Xunzi is the tradition's great sceptic, nobody's ally — but does not think the divining gets him what he asks for. He performs it as form, one of the shapes a public life takes. The gentleman treats it as ornament, Xunzi says, and the common people as spiritual. Same instrument, two registers, and the difference is who is holding it.

Two registers, named from the inside, held apart for two thousand years by people who used both.

The Greeks split the same way. Peter Struck follows Plato, Aristotle, the Stoics and the Neoplatonists as they account for divinatory insight and finds them agreeing on almost nothing except where to look — not at the gods but at the organism. What they explain is a surplus: knowledge arriving without passing through any reasoning we can watch ourselves do. And his reviewers are right that the philosophers took nobody with them. The great sanctuaries were still working in the last hours of paganism, and what they did proceeded, by definition, from the gods.

So the debunkers are right about the register they aim at, and have never in a century hit the other one.

Which leaves the question that is actually open. What was the second register doing that was worth three thousand years of institutional effort?

There are three answers with evidence behind them, none requiring the future to be legible, all published in the last fifteen years in unfriendly journals, and each comes with objections that belong beside the finding.

Start with the one that is hardest to argue with, because it is economics.

In January 2013 Steven Levitt put a website on the internet offering to make people's decisions for them. State the choice, and a coin was flipped: heads, make the change; tails, don't. Twenty-two and a half thousand usable tosses came in — quit the job, leave the relationship, move city, propose. Two months later and six months later, he asked how they were doing.

The coin moved people. On the questions that mattered, someone told to change was about eleven percentage points likelier to have changed than someone told to stay — odd, given that a coin carries no information about whether you should leave your job. (Across all questions the gap is nearer twenty-five points, but that figure is carried by diet and bad habits.)

Then the finding. At two months, nothing: the causal estimate on happiness is indistinguishable from zero. At six months, on the important questions, those who made the change were better off by two points on a ten-point scale — an enormous effect for this measure, and significant. Using the toss as an instrument, the estimate survives the objection that people who make changes were different to begin with.

Levitt's conclusion is not that coins are wise. It is that we are measurably too reluctant to change things: where we genuinely cannot tell, the status quo wins more often than it should, and a randomizer corrects the bias by not having it. He also writes, in the same paragraph, that the results are "merely suggestive." Keep both sentences.

The problems are real and most are his own. The sample is self-selected — young, male, educated, willing to hand a decision to a stranger's website — and the estimate holds only for people marginal enough to let a coin decide. A third of tosses produced no follow-up, and happiness is self-reported by people who have just made a hard change.

The sharpest problem is one he measured himself. The site told participants he hoped they would follow the coin; those who did were several points likelier to answer the follow-up, and he estimates up to a third of the coin's apparent effect may be that artefact.

One more complication, and this essay has to be honest about it because it cuts in the essay's favour. A separate experiment finds that the act of flipping intensifies your affective reaction — that watching the coin land tells you what you were hoping for. If so, the coin is moving people through a channel other than the action it recommended, which is a real problem for Levitt's identification.

It is also, more or less, what everyone who has ever flipped a coin already knows. Hold that thought.

The second answer explains something the first one leaves open, which is why anybody needed an apparatus. If the whole benefit is breaking a tie, a coin would do. Why the stalks, the memorized corpus, the trained interpreter, the years?

Pascal Boyer's answer turns on a piece of ordinary cognitive machinery that is usually invisible because it works so well.

We are extremely good at discounting information by its source. It is among the most reliable findings about human reasoning, and not a flaw: an organism that took all advice at face value would be a catastrophe. When your friend tells you to leave him, part of you is already computing what she gains by it, what she thinks of him, what she needs to believe about her own marriage. The discount happens whether you want it or not.

An answer produced by a procedure has nothing to discount.

Boyer calls it ostensive detachment, and his claim is carefully negative. The procedure does not give you a reason to believe the answer; it removes the reasons you would otherwise have to doubt it. Nobody chose the result and nobody can be suspected of having chosen it, and the machinery around a divination — the randomizer the practitioner does not control, the public procedure, the interpreter who was not told the question — is aimed at exactly that.

Which is why advice fails in the cases where people reach for this. Hard decisions are hard because everyone near you has a position, and an unauthored answer is the only kind you hear cleanly.

There is a second half to Boyer's argument that summaries drop, and dropping it misrepresents him. In a small community, divination is how a group reaches a verdict nobody can be blamed for. Any human-authored diagnosis — this man is the thief, this is why the child died — exposes its author to everything the diagnosis costs the people it names. A procedure produces the same content with no author to retaliate against. That is why it became institutional rather than private, and why it clusters around the decisions that would otherwise tear a village apart.

Now the objections, which in this case are printed on the facing pages: the journal publishes its major articles with commentary attached and the author's reply after it.

The best of them goes at my framing rather than Boyer's. If detachment is the point, the perfect instrument is a coin — no interpreter, no corpus, nothing to distort. Manvir Singh went to the ethnographic database and counted. Across three hundred and fifteen societies, coin-flipping as a divinatory method appears in a single paragraph. Entrail-reading appears in thirty-eight. Bird augury in sixty-nine. Spirit possession — the method most wide open to the practitioner's own influence — in five hundred and thirty-nine.

People do not reach for maximum detachment. They reach for methods that leave a great deal of room for a human being.

Boyer concedes it: detachment is necessary, not sufficient. The concession is the most useful thing in the exchange, because it says what the apparatus was for. A coin gives you a decision and nothing else. What the stalks and the memorized corpus add is content — a body of situations and the ways they turn, to read your own against. Detachment gets an answer past your defences. The corpus is what the answer is made of.

He gives ground elsewhere: Hugo Mercier argues that removing doubt cannot by itself produce belief, and Olivier Morin that nothing stops a dissatisfied party consulting again. Boyer withdraws the strongest version of his claim — the model, he writes, predicts adhesion but not persuasion. People behave as though the verdict settled it; whether they believe it is another question.

The paper is a synthesis rather than an experiment, and its central mechanism has not, as far as I can find, been tested on anybody. Take it as the best available explanation of a pattern, which is what it claims to be.

The third answer comes from someone who has spent decades sitting with actual diviners rather than theorising about them.

David Zeitlyn has worked among the Mambila, who live mostly in Nigeria and partly in Cameroon, since 1985, on a divination whose instrument is a spider. A burrow is ringed with a pot whose base has been knocked out; the hole is covered with incised leaf cards; a stick and a stone sit either side. A question is put in yes-or-no form — stick for one answer, stone for the other — the enclosure is tapped, and the spider comes out and moves the cards. Where they settle is the answer. Zeitlyn is an initiated practitioner himself, and his corpus runs past six hundred cases.

Two things there are worth having.

The first is how much work goes into the question. Getting a real trouble into a form a spider can answer is neither quick nor trivial, and it is done jointly, out loud — a client asking about an American election had the thing built into a proposition before it could be put: the old president, the stick; a new one, the stone. What a session produces is as much a well-formed question as an answer.

The second cuts deeper. Zeitlyn's argument is about how a prediction gets evaluated once people can act on it. Take the advice and you change the situation, so the prediction is falsified — and inside the practice that counts as confirmation, because avoiding the thing was the point. Which is why he ends where he does: the test of a divination is not whether it was correct but whether it helped. In a world of counterfactual conditionals, he writes, the diviner and not the philosopher is king.

His own numbers keep this essay honest. Of a hundred Mambila sessions logged in 2008, sixty-six were diagnostic and thirty-one predictive; in Evans-Pritchard's Zande material the split is nearer even. Both registers, one practice — which is why Zeitlyn says logics, plural.

Now the part that has to be handled carefully, because it is where this subject usually loses its credibility.

Jung did not stop at describing his experience. He proposed a mechanism and called it synchronicity: meaningful coincidence as a feature of nature rather than of the observer.

It is not a scientific claim, and Pauli — whose name is always recruited to prop it up — told a colleague privately that such phenomena, unlike the acausalities of physics, elude capture in law precisely because they are not reproducible. Roderick Main, who has written the most careful account of the concept and whose career sits inside Jungian institutions, gives the verdict worth having: synchronicity has not been disproven or discredited. It has been disregarded — still rarely used by any researcher not already self-consciously Jungian.

Keep the experience and the metaphysics apart and both survive. That a text answered him, and aptly, is real and worth explaining; the three mechanisms above explain it. The principle on top is not evidence, and defending it is the surest way to get the subject dismissed.

Set the three mechanisms side by side and something becomes obvious.

The coin corrects a bias in the person flipping it. The unauthored answer defeats a discount applied by the person receiving it. The question-framing operates on the person struggling to say what they want to know. And the coordination Boyer describes — the part that made these things institutions rather than private habits — works between people rather than inside one.

Not one of them requires the world to be readable. Every one requires a person to be present, in difficulty, with something at stake. These are instruments pointed at us rather than at the future.

Which is what the second register was always for, in the sentence the modern argument keeps not hearing.

Not predictions. Not instructions. Seeing.

The word does exact work. Prediction is a claim about the future, and these systems do not support one. Instruction is a claim about authority, and the moment an answer carries authority it acquires an author and the mechanism collapses. What is left when you take away both is a way of looking at your own situation from a position that is not yours — built out of stalks and spiders and cast coins because a person cannot simply decide to stand outside their own life. Somebody has to be handed a vantage they could not have taken.

Go back to the twenty thousand coin flips.

Every one of those people already knew something. That is what put them on the website. Nobody flips a coin about a decision they have no feeling for; the ones who came had been circling a thing for months, could argue both sides fluently, and had somehow not been able to move.

The experiment about affective reactions was measuring that from the other side — the half-second after the coin lands, when you discover what you had been hoping it would say.

The coin did not know whether they should leave. What it removed was the last thing between them and a decision they had already, in some unfinished way, made: the need for it to be theirs, and defensible, and safe from whoever would ask why.

It was not right about the world. It was right about them.

And it could only be right about them because it knew nothing at all.