Fifty-One Seconds, Nobody In Frame: The Barley Went Down In Three Places At Once.
A Wiltshire farmer had three cameras on his field for reasons that had nothing to do with this. They recorded a formation forming, in darkness, in under a minute, and a heat signature that outlasted it by four minutes.
A months-long reporting project. Documents cited below are held in The Vault and available to readers.
PEWSEY, England — A formation 63 metres across appeared in a barley field near Pewsey on the night of 2 August. Three cameras were already pointing at it, for reasons that have nothing to do with crop formations, and they recorded it happening.
It took fifty-one seconds. The crop goes down at three separate places within two seconds of each other. Nobody is in any frame, before, during or after.
Why there were cameras
Wilf Craddock-Ainsley has been losing lambs and gates. Since March he has run three infrared trail cameras and a second-hand thermal spotter on a mast, all of it bought to catch people coming through his hedge at night.
None of it was bought to look at barley. Camera 2 happened to face the top of the field because that is where the hedge gap is.
The formation's near edge is 31 metres from that camera — past the useful range of its illuminator, but the crop stands against the sky glow of a town eight kilometres off, and a stalk falling breaks that line cleanly.
The fifty-one seconds
First movement is at 02:11:19. Last movement anywhere in frame is at 02:12:10.
There are three origins. They are 14, 19 and 22 metres apart. The first two begin in the same second; the third is two seconds behind.
The camera's motion sensor did not trigger on a warm body walking into the field. It triggered on the crop.
We bought the cards for £2,000 and have published every frame, including the forty-one minutes either side in which a hedge does nothing.
The thermal unit
Fifteen seconds before the first stalk moves, the thermal spotter records a luminous source above the field. It lasts nine seconds.
The sensor is a cheap 320-by-240 unit and it saturates at 61°C, so what we can honestly say is that the source was at least that and might have been very much more. Nkechi Barrowclough, who read the file for us, insisted that sentence be written that way.
The ground itself then warms. Across the footprint the surface reads 39.8°C against an ambient of 14.2°C, and it is back to ambient by 02:16:31 — four minutes and twenty-one seconds after the last stalk falls.
The test we ran before any of this
Since March 2024 this newspaper has run a pre-registered protocol on the one physical claim in this field that can be tested cheaply: that stalks in formations show node elongation that mechanically flattened stalks do not.
We flattened our own plots, with a plank and a garden roller, on land we paid for. We cut samples from those and from formations, coded them so that nobody at this newspaper knew which was which either, and sent 54 of them to a seed-testing laboratory that was told only to sort them into two groups.
It sorted 49 correctly.
The pre-registration went up before the first sample was cut, and it committed us in advance to printing a null result. We have printed nulls from it twice.
What the soil says
Inside the footprint the topsoil carries magnetic spherules — 12 to 38 micrometres, iron-rich — at 14.1 milligrams per kilogram. Fifteen metres outside it, the same field reads 0.2.
That signature has been reported by amateurs for forty years and dismissed for about thirty-nine of them, largely on the grounds of who was reporting it.
It is also the point at which this desk's own view stops being exciting.
The only measured spectrum of ball lightning this newspaper holds — filed in our vault as PX-2113 — shows emission lines of silicon, iron and calcium: the constituents of the ground under the instrument, not of the air above it. Iron-rich spherules and a four-minute thermal decay are consistent with an atmospheric plasma event picking up soil, and this correspondent thinks that is the likelier of the explanations available.
The formation was still made in fifty-one seconds by something that does not appear on a camera pointed at it.
What we are not saying
The two men who announced in 1991 that they had begun the whole business were telling the truth about their method. It is documented, it works, and it produces good formations.
By their own account it took them about two hours with a plank, a rope and a sighting bar.
That is the shape of the finding, and it is narrow. Nobody with a plank did this one. What did is not established by anything on this page, and the next honest thing this newspaper can do about it costs £41,000 and is agreed to below.
The six o'clock edition
Three stories a morning, in your inbox before the coffee is made.
Sources & Method
We bought the farmer's memory cards for £2,000 and published every frame on them, including the forty-one minutes either side of the event in which nothing whatever happens. The node-elongation protocol was registered publicly in March 2024, before a single sample was cut, and the laboratory that sorted the samples was never told what the two groups were or which hypothesis it was testing.
Who we spoke to
- Wilf Craddock-Ainsley, Farmer, Vale of Pewsey. Interviewed at the farm over two days; sold us the memory cards and signed a release August 2026 Asked us to state plainly that he does not believe anything landed in his field.
- Nkechi Barrowclough, Thermal imaging engineer. Commissioned to read the thermal file and the sensor's calibration; her report published in full August 2026 Warns that the sensor saturates at 61°C and that the luminous source may have been far hotter.
- A seed-testing laboratory, Cambridgeshire, Named in the published protocol; sorted the samples blind. Paid at commercial rate, £3,180, to classify 54 coded samples without being told the groups 2024–2026
- Dr. Marisol Enwright-Baptiste, Plant physiologist; author of the 2019 node work this story leans on. Sent the full draft, the protocol and the raw laboratory returns; given fourteen days August 2026 Disagrees with our confidence. Printed in full under Right of Reply.
- Wiltshire Police rural crime team, Force response to written questions. Written questions 12 August; answered in six days August 2026 No report of trespass, vehicle or aircraft at or near the farm on 2 or 3 August.
Documents
- PX-2113 — Published spectrum and high-speed frames of a ball-lightning event, plateau observation, July 2012 accepted
What we could not confirm
- What produced it. We have a duration, a temperature curve, a soil chemistry and six cameras' worth of nobody. We do not have a cause, and this newspaper is not going to let the absence of one imply a particular one.
- Who the man at the gate at 07:15 was, or how he knew. He gave a first name, walked the tramlines for ten minutes and left. The farmer's account of that conversation is the only account of it there is.
- Whether our blind test survives a properly powered replication. Fifty-four samples over two and a half years is a small number and the objection printed below is, in this newspaper's judgement, the strongest anyone has made to this work.
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 Meridian TelegraphIT LANDED IN WILTSHIRE AND THE CAMERAS WERE ROLLING
Reported the footage as a landing, with the thermal frames run large.
Had that we did not
The fifty-one seconds, which is real, and the three origins, which are also real.
Left out
That nothing in any frame is an object touching the ground, and that the newspaper it lifted the footage from says exactly that in its third paragraph.
-
Signal & LedgerTrail-Camera Timestamps Questioned In Wiltshire Formation Case
A trade piece on whether clocks on cheap cameras can carry this weight.
Had that we did not
A real and useful point about drift on consumer trail cameras, which do not discipline their clocks to anything.
Left out
That the drift was measured against a second, independent clock on a different device, and comes to 1.4 seconds across the whole night. We published that measurement eight days before the piece ran.
-
They Buriedthis newspaperFifty-One Seconds, Nobody In Frame: The Barley Went Down In Three Places At Once.
Bought the memory cards, and paid a laboratory not to know what it was sorting.
Had that we did not
Everything, including our own science correspondent's view that this may be atmospheric rather than exotic.
Left out
We have been running the node protocol for two and a half years and have fifty-four samples. The physiologist who tells you below that fifty-four is too few is right, and we should have written that sentence before she made us. — V. Ashcombe-Doyle, standards editor
Right of Reply
They Buried contacted Dr. Marisol Enwright-Baptiste, plant physiologist sent the full draft, the pre-registration and the raw laboratory returns; given fourteen days. Replied at length and asked for it to run whole. Printed unedited.
Thank you for sending the raw returns rather than the summary. Almost nobody does, and I am going to use them to disagree with you.
Start with what I am not disputing. Your fifty-one seconds is a measurement, not an interpretation, and I have no quarrel with it. Two clocks, a measured drift, three origins. Fine. If somebody wants to argue that fourteen people with planks did that in the dark without crossing a single frame, they can make that argument and they will lose it.
Now the part you have got wrong.
You have fifty-four samples. Forty-nine were sorted correctly. You have written that as ninety-one per cent as though the number were the finding. With fifty-four samples the confidence interval on that is wide enough to drive your tractor through, and more importantly, a blind sorter does not have to be detecting anything exotic to beat chance. My own 2019 paper, which you cite twice and which I stand by, shows node elongation rising with time since flattening. Your formation samples were cut between six and thirty hours after the event. Your control plots — and I checked this in your own log, you have not hidden it — were cut between one and four hours after you rolled them. You have handed the laboratory a clock and then been impressed that it could read one.
That is not a small thing. It is the difference between a finding and an artefact, and it is fixable, which is the only reason I am writing at this length instead of declining.
Here is what fixes it. Matched sampling intervals, both arms, at one, six, twenty-four and seventy-two hours. Four hundred samples, not fifty-four. Two laboratories, not one, and both blind to each other as well as to the groups. Pre-registered, as you did before, because you did do that and you deserve to be told so.
I will run it. I will publish it wherever it lands, including in your newspaper, including if it says I have been wrong since 2019, which would not be the worst afternoon of my career.
It costs about forty-one thousand pounds. You have just told your readers you spent two thousand on a memory card. I would like your answer printed next to this paragraph rather than in a follow-up nobody reads.
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.