One Microfilm Reel Was Loaded Backwards. Eleven Thousand Datelines Moved.
The catalogue is fixed; the citations are not. Forty-one published works and three court filings still date pages by a 1987 reel's frame order rather than by the date printed at the top of the page.
BOSTON SPA, England — There are 11,204 pages in the national newspaper archive whose catalogue date does not match the date printed at the top of the page. Some of them appear to report events up to nine days before those events happened.
They do not. A reel was loaded the wrong way round in August 1987.
What the 2019 pipeline was told
The library re-processed 1.4 million pages through new optical character recognition in 2019. The pipeline carried a rule, numbered 7 in its own specification, for the case where the masthead date on a page cannot be machine-read: take the date from the reel's sequence instead.
The rule is sensible, and for 1.4 million pages it is right. A reel runs in date order. If the frame before is the fourth and the frame after is the sixth, the frame between them is the fifth.
On reel 214/9 the frame before is the sixth.
Reel 214/9 is also silvered, and that is the other half of the fault. Of its 11,860 frames the pipeline could read a printed masthead on 656. The remaining 11,204 had no date on them a machine could see, so every one of them was dated by position — from a reel running the wrong way. The 656 legible frames are what holds the drift down to nine days at its widest; without them it would be five weeks.
The contractor's invoice for that reel, which this newspaper bought from a liquidator's document agent, bills 11,860 exposed frames, the spool closed on 11 August 1987. The count is correct. Nobody was paid for a frame they did not shoot. The frames simply run from the end of the run to the beginning of it, and for thirty-two years that did not matter, because until 2019 nothing read the reel but people, and people read the date on the page.
Eleven weeks
Beatrix Sallow-Okonjo of Dundee was looking for her grandfather's inquest. She found a report of a factory fire catalogued two days before the fire, and wrote to say so.
The library's incident file, released to this newspaper under freedom of information after a refusal and an internal review, dates her email to 3 October 2019 and the corrected catalogue to 19 December. Eleven weeks. The file is 41 pages and it does not flatter anybody, which is one reason we believe it.
What the correction could not reach
The catalogue is right now. The citations are not.
Forty-one published works — 34 authors, nine journals, one doctoral thesis — cite pages by the dates the catalogue carried between the 2019 reprocessing and the correction that December, and three court filings do the same. Nothing recalls a citation. There is no mechanism anywhere that goes back into a printed footnote and moves a date by six days.
Professor Hollis Ackroyd-Nwosu, whose 2021 book on the interwar labour press cites eleven of the affected pages, audited his own footnotes at our request and reported back in nine days. Nine of the eleven are decorative. Two carry an argument about how fast a strike spread, and on those two he now thinks the spread was slower than he wrote.
He has published a note saying so. He is the only one of the thirty-four who has.
What this desk takes from it
There is no suppression here, no edit, no removal, and nobody to be angry with. A contractor paid by the frame put a reel on a spindle the wrong way round in 1987. Thirty-two years later a machine was told that order is more reliable than text, and for 11,204 pages it was not.
The nearest thing in this newspaper's own vault, PX-2011, is a chain of printed sources for a claim about giants, and its lesson is the same in a duller register: a mistake in a catalogue can be fixed by one person in eleven weeks, and a mistake in a citation has to be fixed by everybody who ever repeated it, one at a time, for as long as the books last.
Dr. Iseult Farraday-Mbeki's reply is printed below in full. She takes our headline away from us, and she is right that nothing on any page moved, and we have left the headline standing with her objection under it rather than change it quietly.
She also says the library found the fault, fixed it and told nobody, and that this was the only deliberate wrong in the whole affair. We would not have written that sentence about her. She wrote it about herself.
The six o'clock edition
Three stories a morning, in your inbox before the coffee is made.
Sources & Method
We did not set out to check the archive; a reader's email did that, and the library's incident file quotes it. We asked for the file, bought the 1987 contractor's invoices from a liquidator's agent so that we could count frames against the library's account rather than accept it, and then asked one affected historian to audit his own footnotes.
Who we spoke to
- Newspaper Digitisation Programme incident file NDP/19/0447, The library's own investigation of the dating fault — 41 pages, two appendices. Obtained under freedom of information after an initial refusal and an internal review Requested January 2026, released 14 May 2026 Contains the reconciliation that produced the figure 11,204. The figure is the library's. We have not improved on it.
- Bramhall Micrographics Ltd — filming schedules and frame-count invoices, 1986–89, The contractor that shot the tranche, paid per exposed frame. Bought from the liquidator's document agent for £340 — the paperwork, not the negatives March 2026 Reel 214/9 is invoiced at 11,860 frames, the spool closed on 11 August 1987. The count is right. Only the order is wrong.
- Dr. Iseult Farraday-Mbeki, Head of digital preservation at the library. Interviewed by video call twice, then sent the full draft with three weeks to answer June and July 2026 Her reply is printed at the foot of this item and is harder on our headline than on her own programme.
- Prof. Hollis Ackroyd-Nwosu, Historian of the interwar labour press; his 2021 book cites eleven affected pages. Interviewed by telephone, then sent the page images beside the corrected dates July 2026 Audited his own footnotes at our request and reported back in nine days. Two of the eleven matter to his argument. Nine do not.
- Beatrix Sallow-Okonjo, Reader, Dundee; wrote the email that started the eleven weeks. Interviewed by telephone; supplied her original message and the library's acknowledgement June 2026 She was not auditing anything. She was looking for her grandfather's inquest.
Documents
- PX-2011 — Chain of printed sources for the Lovelock 'giants', 1911–2024 accepted
What we could not confirm
- Whether 11,204 is the whole of it. The reconciliation compares the catalogue date against a machine reading of the masthead. Where the masthead cannot be read in either pass there is nothing to compare, and 3,388 pages fall into that class. They are not in the figure and they are not known to be right.
- Whether the reel was loaded backwards or the bound volume was shelved backwards and filmed as found. The frame counts fit both equally. The camera operator named on the 1987 schedule died in 2011, and his logbook entry for that day is four words long.
- Whether any of the forty-one published works is wrong in a way that matters. We asked one author to check and he did, in nine days, and published his own note. Of the other thirty-three we wrote to, eleven have answered. We have not read the books.
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 Library That Quietly Changed Eleven Thousand Dates
Took the correction as the event, and the absence of an announcement as the motive.
Had that we did not
The freedom-of-information refusal and the internal review that overturned it, both real and both worth reporting.
Left out
The eleven-week chronology, which is on page nine of the same file they quote from on page one.
-
Signal & LedgerMetadata Precedence At Scale: One Rule, 1.4 Million Pages
Read it as a pipeline-design lesson for anyone running bulk recognition, and read it well.
Had that we did not
Rule 7 set out in full, more clearly than we have set it out.
Left out
Beatrix Sallow-Okonjo, who is the only reason anybody knows, and who is not a systems problem.
-
They Buriedthis newspaperOne Microfilm Reel Was Loaded Backwards. Eleven Thousand Datelines Moved.
Asked for the incident file, bought the 1987 frame counts to check the library's account against the contractor's, and had one affected historian audit his own footnotes.
Had that we did not
The library's figure, its chronology and the rule that produced the fault, all as the library itself states them.
Left out
We ran a headline about eleven thousand datelines moving when nothing on any page moved, and our own subject had to point it out. We have kept the headline. — V. Ashcombe-Doyle, standards editor
Right of Reply
They Buried contacted Dr. Iseult Farraday-Mbeki, head of digital preservation on 26 June 2026 by video call, again on 3 July with the full draft, the figure, the rule and the frame-count invoices, and three weeks to answer. Replied in writing on 20 July 2026, asking that it run entire. Printed unedited.
Your headline says eleven thousand datelines moved. Nothing moved.
I want to be precise about this because precision is the entire subject of your item. Not one mark on one page was altered, added or removed. Every page in this archive is exactly what it was when it came off the press, and the microfilm is exactly what the contractor shot in 1987. What changed was a number in a database that did not exist until 2019, generated by a machine, about a page it could not read. That is a metadata fault. Describing it as movement in an archive is how a library ends up explaining itself to a select committee, and I would ask you to consider that before you write the next one.
On Rule 7 I will not apologise at all. You have set it out fairly and then let your readers assume it was a blunder. Give me the alternative. Forty-six thousand pages in this archive arrive with no legible date. I can leave them undated, in which case they are unfindable, and an unfindable page in a national collection is functionally destroyed. Or I can date them by their position in a reel, which is right everywhere else in 1.4 million pages and wrong for 11,204, and which is correctable, because the error leaves a trace and the trace is a date you can check against the page. I chose findable. I would choose findable again tomorrow, and so would you if the alternative were sitting in your reading room.
Your citation paragraph is the one I object to most, because it is aimed at us and it does not belong to us. We have no mechanism to recall a footnote. Neither does a journal, neither does a court, and neither does your newspaper. You are describing a permanent property of scholarship — that a claim, once printed, outlives the source it rests on — and you have arranged the sentence so that it reads as a failure of this library. Forty-one works cite a date we published and then corrected. We corrected it in eleven weeks from a standing start, on the word of one reader, without being made to. I have worked in this sector for twenty-two years and I can tell you that eleven weeks is not a scandal. It is the fastest anything of this kind has ever been done here.
Now the thing you are right about, and I will say it plainly because you will otherwise think I have not understood you.
We told nobody. We found it, we fixed it, we filed the report and we published nothing, and the only reason there is a public account of this at all is that you asked for the file and then asked again after we refused. That refusal was wrong. I did not make it and I did not appeal it and both of those are also things I did. If we had put four paragraphs on the programme's news page in December 2019, the thirty-four authors would have had six years to check their footnotes instead of hearing it from a reporter, and Professor Ackroyd-Nwosu would not have had to spend nine days of his summer doing our correction for us.
Print that. Do not soften it, and do not put it after your own conclusion, because it is the only part of this in which somebody did something wrong on purpose.
And one last thing, for Mrs Sallow-Okonjo, who you will quote in a sentence and move on from. She sent us a polite email about a date that looked odd to her, and she was right, and she was right about eleven thousand two hundred and four pages. Nobody in my building found this. A reader in Dundee looking for her grandfather found it. That is what a reading public is for and I would rather your item ended on her than on us.
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.