Nobody Cut The Sentence. She Finished It And The Column Ran Out.
Sixty incident narratives, obtained twice: once as the database export, once as photocopies of the paper originals. Fifty-seven of the sixty run on past the cut. The famous fragment ends with a supervisor being informed at 14:20.
Background written by the news staff. Contains no argument or recommendation.
CROYDON, England — The sentence at the centre of this ends: "the tube slipped and struck the cabinet sash; no breach of glove; supervisor informed at 14:20."
It has been sitting in full, on paper, in a filing cabinet in Croydon, since March 2021. Nothing was ever taken out of it.
What circulates instead is a fragment. A laboratory incident report, released under freedom of information, whose narrative stops dead: "while transferring the second rack the". We counted roughly 9,000 reproductions of it online. It is read, almost everywhere it appears, as a cut.
We ordered the same reports twice
This newspaper asked for 60 incident reports in two ways, in two applications eight weeks apart, so that the two disclosures were prepared by different hands.
The first request asked for what agencies normally release: the database export. The second asked for photocopies of the scanned paper originals of the same 60 reports, and cost £186 in copying charges.
Fifty-seven of the 60 run on past the point where the export stops. The narratives average 61 words, or 337 characters. The three that end at the cut end there because the person writing had finished.
Two hundred and fifty
The column is called NARR_TEXT. It is declared VARCHAR(250) in a schema written in 1998, and the change log shows it has never been widened through four front ends and two hosting migrations.
When a records clerk keys the paper form, typing simply stops when the field fills. No warning is shown, no character counter appears, and nothing in the release marks the entry as incomplete.
Piotr Enwezor-Hale has maintained that schema since 2004 and has answered this question in writing eleven times since 2016 — to five requesters and one parliamentary committee.
Fourteen boxes
We had a blank form annotated twice, separately: by Winifred Lascelles-Obi, who trained on it in 1999 and taught it for eleven years, and by Enwezor-Hale, against the table definitions.
Box 1 is the date. Box 4 is the organism. Box 9 is the narrative, headed "in your own words, for your supervisor". Box 10 is immediate action taken. Box 11 is the supervisor's name and signature.
Between Box 10 and Box 11, on the paper, there is a gap of about an inch.
Both annotators drew an arrow into that gap. Neither knew the other had. One wrote "outcome?" and the other wrote "this is what everybody thinks Box 11 is".
The box that is not there
There is no outcome field anywhere on this form. Not a diagnosis, not a referral, not a yes or a no.
Outcome goes to the occupational-health service, which runs a separate system with no key back to this one. The link was proposed in 2011 and refused on medical-confidentiality grounds, in a signed note the service supplied to us unprompted.
The refusal is defensible and Lascelles-Obi, who was in the room, defends it below at length and with some heat.
But the consequence is flat and worth stating plainly: no incident report in this system can tell anyone whether a person became ill.
What is actually being counted
The register these reports feed is quoted, in submissions and in headlines, as though a rising line meant rising infections. We looked at nine years of that same register separately, and found the rise was a definition.
This is the other half of it. Every row in that register is an event: a spill, a near-miss, a glove. It is not a case count, it has never been a case count, and the field that would make it one was left off the form on purpose by people who thought a technician should be able to report a fright without generating a medical record for her employer to read.
They may have been right. This desk thinks they were probably right.
What follows from it is that the most-quoted number in this subject cannot answer the only question anybody actually asks about it — and that the fix Lascelles-Obi proposes in her reply, a three-word tick box carrying no medical information at all, has never once been requested by anybody, including her.
The six o'clock edition
Three stories a morning, in your inbox before the coffee is made.
Sources & Method
We asked for the same 60 reports twice, in two formats, eight weeks apart, and paid the copying charge for the paper so that the two releases could be set side by side. Then we asked two people who have never met to annotate the same blank form and to say, of each of its 14 boxes, what it is for and what it cannot hold.
Who we spoke to
- Sixty incident reports, obtained twice, The standard database export, and photocopies of the scanned paper originals of the same 60. Requested under freedom of information in two separate applications, eight weeks apart, so that the two releases were prepared independently February to June 2026 £186 in copying charges for the paper. Two of the 60 have a name blacked out; nothing else is withheld on any of them.
- Data dictionary and change log for the incident database, 1998 to 2026, Table definitions, column widths and every schema change in 28 years. Obtained under freedom of information; released in full and without redaction April 2026 NARR_TEXT has been VARCHAR(250) since the first release. It has survived four front ends and two hosting migrations.
- Winifred Lascelles-Obi, Retired biological safety officer; trained on this form in 1999 and taught it for eleven years. Annotated a blank form, box by box, in her own hand, without being told what the other annotator had written May 2026 Her annotation is published entire. Her reply to this story is at the foot of the page.
- Piotr Enwezor-Hale, Database administrator; has maintained the schema since 2004. Annotated the same blank form against the table definitions, separately, and answered follow-up questions by email May and June 2026 Has answered the truncation question in writing eleven times since 2016, to five requesters and one committee.
- The occupational-health service, Written answer on linkage, with the 2011 decision note it refers to. Put to them in writing; answered in writing, and the decision note supplied unprompted July 2026 A link between the two systems was proposed in 2011 and refused on medical-confidentiality grounds. The refusal is signed.
What we could not confirm
- Whether the paper original is itself always complete. Box 9 is a physical box about four inches deep. Six of our 60 continue into the margin or onto the reverse, and one carries an arrow to a second sheet which was not in the file we were sent and which nobody has been able to find for us.
- Whether anything is ever withheld before the scanning stage. We asked for 60 reports and received 60. A report held back in its entirety, or an incident for which no form was ever raised, would look identical to us from here, and there is no independent list of incidents to check the file against.
- Whether the 2011 refusal to link the two systems is still a decision or has become a habit. The note is signed and reasoned. Everyone who signed it has left, nobody has revisited it in fifteen years, and 'we looked at that in 2011' was offered to us twice as though it were an answer.
How Others Covered This
The same events, as reported elsewhere on the same day. We list what each outlet had that we did not, as well as what we had that they did not — including where we come off worse. Why we print this.
-
The Hollow PostTHE SENTENCE THEY CUT: What The Agency Will Not Let You Read
Treated the truncation as redaction and counted the missing words as a measure of what is being kept back.
Had that we did not
The fragment, quoted exactly, and the sound observation that the cut always falls in the same place.
Left out
That the same reports are released complete, on paper, to anyone who asks for the scan and pays 12p a page.
-
ApertureMid-Sentence
An essay on the interrupted record, and on what a reader does when handed a line that stops.
Had that we did not
The best sentence anyone has written on this: that a cut sentence recruits the reader to finish it.
Left out
The 250. One telephone call to the database administrator, who has answered the question in writing eleven times.
-
They Buriedthis newspaperNobody Cut The Sentence. She Finished It And The Column Ran Out.
Ordered the same 60 reports in two formats and had the blank form annotated, separately, by the officer who trained on it and the man who maintains its schema.
Had that we did not
All 60 narratives in both forms, the data dictionary, and both annotations of all 14 boxes.
Left out
Our headline puts a technician in the third person and her sentence in a newspaper, to make a point about a database column she has never heard of. She wrote it for one named supervisor. The finding that matters — that nothing in this system can tell you whether a person got ill — is the fourth section. — V. Ashcombe-Doyle, standards editor
Right of Reply
They Buried contacted Winifred Lascelles-Obi, retired biological safety officer on 19 June 2026 in person in Croydon, again on 28 July with the full draft, both annotations and three weeks to answer. Replied by letter on 14 August 2026. Printed entire and unedited, at her request.
You have got the mechanism right and I want to argue with your manners.
First, the sentence. You have printed in full something a woman wrote in a box that says, at the top, 'in your own words, for your supervisor'. She wrote it in about ninety seconds with her hands still shaking, and she wrote it for one person, whose name is in Box 11 and whose signature is under it. Nine thousand strangers had a fragment of it. You have now given nine thousand strangers all of it, and you did that to correct them, and I accept that the correction is worth something. But do not tell yourself you handed her back her sentence. You published it. Those are different things and only one of them was hers to do.
Second, 1998. You write 'a schema written in 1998 and never migrated' in the tone people use for a joke about a fax machine. I keyed those forms. Two hundred and fifty characters was what the field officers agreed to, because they were being asked to type at a terminal in a corridor, on a shared machine, at the end of a shift, and because the box on the paper was the record and the database was an index to it. The database was never meant to be the thing you read. It has become the thing you read because it is the thing that can be emailed, and nobody decided that.
Third, and this is where I will not give you an inch: you write about the absent outcome field as though the absence were an oversight. It is not. I was in the room in 2011. We were asked to attach a person's health outcome to an incident record that is seen by a line manager, a safety committee, an inspector and, eventually, you. I argued against it and I would argue against it again tomorrow. A woman who reports a glove breach should not be creating a document that tells her employer, four months later, what her blood said. The unlinking is not bureaucracy. It is the only thing on that form that was designed to protect the person filling it in.
What I will concede, because your two annotations caught me out and I have thought about it since, is the arrow. I drew it in the same gap he did, and I did not notice I had drawn it until you told me he had too. There should be something there. Not an outcome. Not a diagnosis. A tick box, three words: 'referred to occupational health, yes or no'. That carries no medical information whatever, and it would mean that a person reading your register could at least tell the difference between an event where somebody was sent to a doctor and an event where somebody mopped a bench.
That box would have cost nothing in 1998. It would cost nothing now. Nobody has ever asked for it, including me, and I taught the form for eleven years.
Published unedited under our right-of-reply guarantee.
How was this story?
We publish the result, whatever it is. Reader verdicts appear on the front page and in our newsroom metrics.
Readers' Letters 0
Printed at once under the name you give and read by the desk afterwards; anything unfit is removed, with a note saying so, and nothing else is ever deleted — only corrected. Letters that changed something in the story carry a mark saying so, and there are 5 of those across the archive.