Going to Look for What Isn't Written── The gap between the material and its source that no one visits

If you have checked everything that is written, why do the errors not go down? The answer is probably simple. Most of them are not hiding inside what is written. They are hiding in what was dropped on the way. However carefully you read the page in front of you, what is not on it cannot be seen.

01I checked everything, and still missed it

I still remember the day I cleared a particular piece. A single-sheet leaflet. I matched the efficacy wording, the footnotes and the axes of the graph one by one against what had been approved. The figures agreed. Nothing reached past the range. No findings. I counted that day as a light one.

About six months later a colleague who had been reading the underlying paper said, almost in passing, "That graph isn't the primary endpoint, is it." I went back to the leaflet. The figures were exactly as the paper had them. Nowhere did it say that this was not the primary endpoint. Since it was not written, nothing within the range I had checked was wrong.

What I had been checking, it turned out, was the consistency of the material with itself. The distance between the paper and the leaflet had been outside my field of view from the start.

Which, on reflection, is obvious. The paper runs to dozens of pages. The leaflet is one sheet. Making one sheet out of it means cutting. Who was enrolled, who was excluded, how many were lost to follow-up, which of several endpoints this one was. What gets cut stays only in the hands of the person who cut it. From the receiving end, what was removed never existed.

02The story closes on the information in front of you

This trap has a name. Daniel Kahneman called it WYSIATI — what you see is all there is. The fast judgment in your head does not ask whether the information is sufficient. It asks only whether the material at hand can be assembled into a coherent story.

The awkward part is that the habit works hardest when there is least to go on. Fewer materials mean fewer contradictions. Fewer contradictions mean a cleaner story. A clean story breeds confidence. A well-organized single sheet is therefore the most dangerous thing on the desk. It holds together, perhaps, because nothing is left in it that could fail to hold together.

Confirmation bias sits on top of that. Ordinarily I set out to check whether the material is right. Everything that supports its being right is inside the material, because that is how it was assembled. The question pointing the other way — if this were not right, what unwritten information would make it so? — does not arise unless you deliberately raise it.

So the feeling of having checked everything that is written is not grounds for reassurance. It reports that you have looked over the whole of your field of view, and says nothing whatever about whether the field of view was the right one.

03Searching under the streetlight

A man who has dropped his keys at night is crawling around under a streetlight. Someone passing asks whether he dropped them there. No, he says, I dropped them over there. But the light is better here. It is an old joke. It does not read as funny from a review desk.

In December 1972 this happened in the sky over Florida. As an airliner came in to land, the light indicating that the nose gear was down failed to come on. Captain, first officer and flight engineer all turned their attention to that light. What had actually failed was a single bulb. The gear was down.

While three people leaned over a small lamp, the autopilot slipped out. The aircraft descended, slowly. The altimeter was directly in front of them. No one was looking at it. The aircraft came down in the swamp, and something over a hundred people died. Everyone had been looking, with great diligence, at the brightest thing in the cockpit.

What happens at a review desk differs in scale but not much in shape. Around a single sheet stand the person who drafted it, the person reviewing it and the person who will use it in the field. All of them are studying that sheet closely. Type size, the placement of footnotes, colour. And in that room, not one person has the underlying paper open.

Nobody was being careless. The opposite: because they were concentrating on what lay in front of them, everything outside it went dark. Diligence does not protect you from this kind of miss.

04"Whose job is that?" and then it is no one's

"Going back to the paper is the drafting side's job, surely." I have heard that many times, and I have probably said it myself. The logic holds. The material was drafted by the drafting side, and the duty to assemble the evidence rests there first. As a division of labour it is correct.

Only, the moment the sentence is spoken, the underlying paper quietly disappears from the reviewer's field of view. What began as a line drawn to make responsibility clear has turned into a reason not to look.

Psychology has an old body of work on this. The more people who witness someone collapse, the less likely any one of them is to step in. It is called diffusion of responsibility: everyone assumes, simultaneously, that someone other than themselves will act. Not one person present means any harm. Still, nobody moves.

In December 1978, above Portland, Oregon, something close to this played out. To work on a landing gear problem, an airliner circled the airport. The captain's attention was full of the gear. The first officer and the flight engineer watched the fuel going down. They said so, more than once, obliquely. They did not say it plainly. The captain did not register it. The aircraft ran out of fuel and came down short of the airport. Ten people died.

After that accident the industry took up CRM in earnest — training that lets a crew say what they have noticed across the ranks, in the moment. The reframing was that accidents are not caused by a shortfall in skill, but by observations that fail to travel between seats.

05What is not in the record does not exist

In October 1966, in a small mining village in Wales, a spoil tip gave way. It was a heap of waste rock built up over years from the coal workings. The slide came down the hillside and engulfed a primary school. One hundred and forty-four people died, one hundred and sixteen of them children.

The inquiry found that the tip had been built over springs. They were marked on the geological maps. Local people knew about them. There had been smaller slips before. And yet, in the National Coal Board's own management records, that tip was a tip with nothing wrong with it. A danger absent from the paperwork was, to the organization, a danger that did not exist.

Japan has an accident of similar shape. In November 1963 a coal dust explosion tore through the Miike mine. Four hundred and fifty-eight people died, and roughly twice that number were left with carbon monoxide poisoning that stayed with them for years. Fine coal dust lying underground was, to the people who worked there, part of the ordinary scenery. Ordinary scenery does not appear on a list of hazards. Not being listed, it is never counted.

James Reason set out accidents this way: beneath the unsafe acts you can see lie latent conditions — weaknesses that have been buried in the design, the procedures and the customs for years. Repair only the visible side and the buried side remains. While it remains, the same accident will one day happen again, at someone else's hands.

Back to the material. What appears on a list of findings is only the kind of error that someone noticed. A divergence occurring in a place no one went to look is recorded nowhere. And what is not recorded does not reach the tally, the report, or the improvement plan. Zero findings is not proof that there was nothing. It may be proof that nobody looked.

06If errors persist, look at the side that is not written

When the same kind of error keeps coming back, we usually add another thing to check. A line is added to the list. The following month something similar happens. Another line is added. And still it does not fall.

It does not fall because everything we are adding sits on the visible side. However finely you subdivide the inside of the material, the distance between the material and the source paper does not become one millimetre more visible. Refining the ruler does not measure the place you are not measuring.

Erik Hollnagel, who studies safety, wrote that the work as imagined and the work as actually done always diverge. Reading the procedure tells you nothing about the floor. It tells you how the procedure was written. The same holds here. Reading the material tells you what the person who made it wanted to convey. It does not tell you what the underlying data say.

So when errors are showing up, the side to look at is not the written one. It is the side that was not written. In practice that comes out like this.

One more thing deserves care here. Diane Vaughan, studying the space shuttle, named the normalization of deviance. Each time you drift a little off the standard and it passes. Before long the drifted state is the normal one. "It went through this way last time" starts to behave as though it were evidence. Last time's judgment was about last time's material, and says nothing at all about this time's source data.

07Doubting is not the same as distrusting

Say "let us go back to the paper" and it can sound as though you doubt the person who drafted it. Sometimes it is taken that way and the room stiffens. So let me be plain about this part.

The gap in information arises without anyone's ill will. Compress dozens of pages into one sheet and something must be lost. The decisions about what to lose are usually made in good faith: make it readable, get the important part across. Every one of those good-faith decisions becomes, from the receiving end, a hole you cannot see. The hole opens not because people are dishonest, but because that is what compression is.

Going back to the source, then, is not a statement about trust. It is done because you know that compression always drops something. That is not an assessment of a person; it is an understanding of what the work is. Since I started thinking of it that way, saying "shall we look at the paper together?" has become a great deal easier.

And this is also a question of where the conscience sits. Checking what is written produces a record that the duty was discharged. Going to look at what is not written leaves no record. Nobody praises it. Skip it and the day ends without incident. Whether you go and look anyway is settled not by the procedure, but by what a person actually cares about.

The ability to look carefully at what is visible is the foundation of this work. That is not in question. But it is equally clear that there are places it cannot reach. While three people leaned over a light bulb, nobody was watching the altimeter. The tip that was safe on paper had been dangerous on the geological map for years.

What we miss is rarely what we declined to look at. It is what we never counted as something to look at in the first place. So if you want fewer errors, before you narrow your eyes, check once where the edge of your field of view is. What is not there will stay unseen however hard you look.