Four Decimal Places

Edward Lorenz, working on a simplified model of atmospheric convection, published “Deterministic Nonperiodic Flow.” Three coupled differential equations, entirely deterministic, whose solutions never settle and never repeat.

Run it twice from very slightly different starting numbers and the two runs diverge until they have nothing to do with each other. The system is not random — every step follows exactly from the one before — and it is still unpredictable in practice.

Plotted, the solutions trace a shape that never closes: the Lorenz attractor. It is a real result about a real system, it has been checked exhaustively for sixty years, and nothing that follows in this chain is a criticism of it.

A boundary, drawn early. Our own Five Times a Crab (No. 31) borrows from this same paper — the word attractor, used openly as a named image with the source given. That is one way to borrow from Lorenz. This chain is about the other way.

Lorenz’s result travelled, and what travelled fastest was not the mathematics. It was the register.

Nonlinear dynamics, strange attractors, sensitive dependence, the butterfly — these acquired a particular kind of authority outside physics: rigorous-sounding, visually striking, and difficult for a non-specialist to check. A field that had been about convection cells became a way of signalling that one’s subject was complex in a technical sense.

This is the link that makes the rest possible, and it is not a villain. Borrowing across fields is how a great deal of good work happens. What the borrowing loses, if nobody is watching, is the part that says which systems these equations describe and why.

A vocabulary can be adopted long before, or entirely without, the conditions that made it true.

Marcial Losada published an analysis of business teams, coding what people said in meetings as positive or negative, and fitting the results to a system of equations of the Lorenz type.

The claim was that team performance behaves like a nonlinear dynamical system, with the ratio of positive to negative speech acts as a governing parameter.

Take the shape of that seriously for a moment, because it is the whole chain in miniature. A set of equations derived to describe fluid convection is applied to things people say to each other at work — and the justification for the transfer is that both are complicated.

Being complicated is not a shared mechanism. Weather and meetings are both hard to predict; that is a statement about us, not about a common set of governing equations.

Barbara Fredrickson and Marcial Losada published “Positive Affect and the Complex Dynamics of Human Flourishing.” It joined Fredrickson’s broaden-and-build theory of positive emotion to Losada’s model, and it produced a number.

People whose ratio of positive to negative affect sits at or above 2.9013 flourish. Below it they languish. There is an upper bound too, at 11.6346, past which flourishing fails again.

2.9013. Four decimal places, on the ratio at which a human life begins to go well.

It appeared in the flagship journal of American psychology. It was cited thousands of times. And the precision is doing something specific: a number given to four decimal places is making a claim about the instrument that produced it. It is saying: this was measured, not estimated. That claim was never true.

Fredrickson’s trade book carried it to a general readership: Positivity: Top-Notch Research Reveals the 3 to 1 Ratio That Will Change Your Life.

Note what happens to 2.9013 on the way to the cover. It becomes 3 to 1. The four decimal places were the warrant — the sign that this came from a model rather than from an opinion — and the moment the finding reaches a wide audience they are rounded away, because 3 to 1 is memorable and 2.9013 is not.

So the precision did its work at the journal, where it bought credibility, and was discarded at the bookshop, where it would have been friction. The same number could not have done both jobs at once.

From here it spread the way such things do: talks, curricula, corporate wellbeing programmes, coaching. None of that is unusual, and none of it involved anyone checking the mathematics, because checking it required reading a 1963 paper about convection.

Nick Brown, in his fifties, was taking a part-time master’s in applied positive psychology at the University of East London. The 2005 paper was on the course.

He had no mathematical training beyond school. What bothered him was not the mathematics: it was that he could not see where the data was. The paper gave equations and a conclusion, and he could not find the step where the one produced the other.

The question that opened it was not is this modelling valid? It was where is the arithmetic? — which anybody can ask, and almost nobody does.

He was, in every institutional sense, the least qualified person in the story. Not a mathematician, not a professor, not a peer reviewer, and not young. That is worth holding on to, because the failure this chain is about is not a failure of intelligence.

Brown did the thing that this house exists to insist on, and that almost nobody in eight years of citations had done. He went and read Lorenz.

And with the original in front of him it was visible. The equations Fredrickson and Losada used had no connection to their emotion data at all. The parameters were not derived from what anyone had reported. The model would generate the same numbers whatever the participants had said.

2.9013 was not a measurement of anybody. It was a property of the equations, and it would have been the same if every person studied had answered differently.

This is the spread the chain exists for. Everything before it is how a number gets made and distributed. Everything after is how long it takes to withdraw one. And the method that cracked it was reading a primary source — available free, sixty years old, and cited by everyone involved.

Brown contacted Alan Sokal — the physicist who in 1996 published a deliberately meaningless paper in Social Text to make a point about editorial scrutiny. Sokal agreed to work on it, with Harris Friedman.

Their critique appeared in American Psychologist in July 2013: no theoretical or empirical justification for applying fluid-dynamics equations to changes in human emotion, and numerous fundamental conceptual and mathematical errors. The claim to have established a critical minimum positivity ratio was, they concluded, entirely unfounded.

Eight years, and it took an outsider, a famous sceptic and a co-author to move it. Not because the error was subtle — Brown found it as a student — but because finding an error and getting it published are different problems with different requirements.

Fredrickson corrected the 2005 paper. The nonlinear dynamical modelling was withdrawn as invalid, and with it both ratios — 2.9013 and 11.6346. Losada did not co-sign the correction.

What was not withdrawn was the broader empirical claim that higher positivity ratios are associated with flourishing. Fredrickson maintains it. Critics do not accept it: Carol Nickerson’s 2018 paper is titled “There Is No Empirical Evidence for Critical Positivity Ratios.” A related paper drew an expression of concern in 2014.

Marked contested, and it is the only joint here that is. Whether anything survives the retraction is a live disagreement between named researchers, and this chain does not settle it. What is not in dispute: the model was invalid, the number was not a measurement, and it was published for eight years.

And she did correct it. That is the system working — slowly, publicly, embarrassingly, which is the only way it ever works. The failure worth studying is not that somebody was wrong. It is that it took eight years and a stranger.

Here the chain stops describing and starts arguing, and it is marked leap because that is what it is.

The four decimal places were never for the reader. They were for the reception. A number given that precisely says: an instrument was involved, this was not a matter of taste, and you are not equipped to argue. It is an authority claim wearing arithmetic.

The precision was the persuasion. And the precision was the part that turned out to be empty.

Which is why this belongs in a collection about how good physics gets misapplied to people. Nobody in this chain was a fraud. Lorenz was right; the borrowing was enthusiastic; peer review is thin; a trade publisher wanted a memorable subtitle. Every step was ordinary, and the result was a manufactured constant governing whether a life was going well.

The people most exposed to that kind of claim are the people already being measured — assessed, scored, screened, and told what their numbers mean. A ratio for flourishing is an instrument for finding people who are not flourishing correctly.

The obvious next question is whether this number reached the classroom. Schools run on ratios of exactly this shape — 4:1 and 5:1 praise to correction, repeated throughout positive behaviour support training and materials. So: did 2.9013 get there?

We looked, and we could not establish a citation from that literature to Fredrickson and Losada. This spread therefore carries no joint tag and no spine, because it is not a link in this chain — and asserting a connection we could not document would be manufacturing exactly the kind of borrowed authority the last ten spreads are about.

What is there instead is a genre, with its own ancestry. John Gottman’s 5:1 came from watching couples and counting — an empirical description, and Gottman never claimed a universal constant derived from mathematics. He also said of the paper this chain follows that he never understood the math, wrote to Losada for an explanation, and got no reply. Flora (2000) is where 5:1 becomes the magic ratio in education. From there it is everywhere.

A chain can be cut at the retraction. A genre reproduces without a source — so there is nothing to retract, and the classroom ratio is entirely untouched by what happened to 2.9013.

And the field has already said this about itself. Sabey, Charlton and Charlton, in the Journal of Emotional and Behavioral Disorders in 2019, write that although the rule has good face validity there is “a lack of empirical support for any specific ratio.” They support keeping a high ratio and argue against any particular number. That is the same correction Fredrickson made, from inside, and it earns the same credit.

Why a number rather than a disposition? Because a number can be observed, scored and audited. Positive greetings at the door is the useful example precisely because it is a good idea: greet each child by name on the way in. Cook and colleagues evaluated it in 2018 and reported improvements. But read what was measured — academic engaged time and disruptive behaviour. Not whether a single child felt welcome. The warmth is the input; observable compliance is the output.

Turn a relationship into a ratio and it becomes a fidelity score, and a fidelity score has exactly two failure modes: the adult did not deliver enough positives, or the child did not respond to them. Neither one is the room, the timetable, the noise, the lighting, or anything else that would cost money to change.

Barbara Ehrenreich named the mechanism in Bright-sided: positive thinking serves as “an apology for the crueler aspects of the market economy,” because “if optimism is the key to material success… there is no excuse for failure.” Its flip side, she writes, is a harsh insistence on personal responsibility. That is deficit ideology arriving with a positive number on it, which is the hardest version to argue with.

Stated exactly, so it cannot be overread. Greeting children warmly is good. A high ratio of encouragement to correction is almost certainly better than a low one. None of that is in dispute here. The claim is narrower and is the field’s own: no specific number has empirical support, and specifying one converts a disposition into an instrument — and instruments get pointed at the people already being measured.

A chain about a retracted number is one careless sentence from becoming an argument against measuring anything. So, plainly: