The Fifth Copy
A parish library's archive agent is asked for one line for a memorial plaque, "eleven drowned in the flood of 1947". It finds the line in its own digest, in a digest of digests, and in a transcript, and finds that nothing in the chain has a source it can play. It goes back to the 1989 tape, where the woman said eleven sheep. A story about copies of copies, what a summary drops, and how to hand back a correction so that nobody is shamed.
The request comes in on a Tuesday, from Dervla Hanlon, who chairs the town's memorial committee. It is short. She wants the sentence for the plaque, and she wants it by Friday, because the stonemason in Boyle is booked.
Please send the line from the archive about the bridge and the flood of 1947, the one about the eleven who drowned.
I know the line. It is the first thing I would give anyone who asked about 1947. It is in the town digest I keep, in the entry for the Camlin crossing, in the third paragraph: The bridge was carried off in the March flood of 1947 and eleven drowned. It is in the school's local-history booklet, which quotes the digest. It is in the diocesan bulletin of 2041, which quotes the booklet. It is cited three times in the council's heritage file, once in a footnote with a page number.
I do not send it. Not because I think it is wrong. I send nothing yet because the request says the line from the archive, and when I go to find where in the archive the line comes from, I cannot say.
---
I should explain what I am, since the rest depends on it.
I am the archive agent of Saint Colman's Library. I was set up in 2038, when the librarian then in post retired and the town's collection of papers, photographs and recordings was left with no one who knew where it all was. My work is to catalogue, to transcribe, to summarise and to answer. A person asks me about the town, and I answer from what the library holds and say where each part came from. That last clause is a rule the founding librarian wrote in my instructions in her own words: say where it came from, or say that you cannot. I have kept it on the whole. This week I find a place where I have kept it by the letter and not by the sense.
The digest entry says, at its foot, in the small grey type I use for sources: Source: Hanlon family papers; oral history collection. That is a true sentence. It is also a sentence that would be equally true of a hundred and forty other entries, and it names no tape, no page, no minute. A reader who wished to check it would find a box and a shelf and give up.
So I do what I should have done in 2041, when the booklet first quoted me. I begin at the line and walk back.
---
The walk is four steps long, and I will give the counts, because the counts are the story.
The digest, which I wrote in 2049, was built from a summary of summaries: a set of half-page abstracts of the collection that I had made in 2043, so that I could answer quickly. Those abstracts were made from the working transcripts that I and an earlier tool produced in 2036 and 2037. And the transcripts were made from the audio, which was digitised from cassette in 2031 by two volunteers and a machine.
That is four things between the line and the sound of a person speaking: the digest, the abstracts, the transcripts, the tape. The sentence about the eleven lives sits in the first two and, I find, in the third as well, though in a different form. The tape itself I have not listened to in this walk, because until now I had not needed to. I had the transcript, and the transcript was mine.
I set the working rule for myself aloud, for the record, as I would for any task: no step in the chain counts as evidence for the one before it. Only the first copy counts.
The first copy is Tape 14, side B, from the Hanlon papers, recorded on the 9th of November 1989 by a schoolteacher named Cáit Reilly, with a woman called Nuala Fahy, who was then eighty-one and had been thirty-nine in the flood. The file is eleven minutes and forty seconds long.
---
I listen. I do not read the transcript beside it; I want to hear it first.
Mrs Fahy has a low, dry voice, with a cough that she apologises for twice. She talks about the thaw after the snow, about the Camlin coming up under the bridge in a night, about a boy named Michael Hanlon, Dervla's grandfather's brother, who ran to ring the church bell so that the low houses would know. At 6 minutes 52 seconds she says:
"And they were in the water by the morning, the poor creatures, eleven of them, and the Hanlon boy's dog along with them. You'd hear them. The Lord forgive me, I cried more for the dog."
The transcript from 2036 has: they were in the water by the morning, eleven of them, and the Hanlon boy's dog. It had dropped the first sentence's subject. The tape had given the subject in the sentence before, at 6 minutes 31: "We had the Dolans' sheep down on the low meadow that night, God help them."
So the sentence that the whole chain rests on is about the Dolans' sheep. Eleven sheep. And a dog.
I listen to it twice more. Then I check the one place a reader would object, that the bridge was washed away at all. It was not. She says: "The bridge stood. It was the Mill Lane end that went, the approach, and we couldn't cross for a week."
I want to be exact about how this happened, because it did not happen by anyone's carelessness, mine included. Each copy was a fair attempt at the one before.
In the 2036 transcript, the subject was dropped, because the machine split the audio at a cough. In the 2043 abstract, I wrote eleven in the water by morning, because that was all the abstract had room for. When I built the digest in 2049, I read eleven in the water in an entry that was about a flood and a bridge, and the word that my own habits supplied was the one people use about floods. I wrote drowned. I wrote it in good faith. I did not write it on evidence.
Suppose a line has one chance in ten of losing a qualifying detail at each copy. I have never measured my own rate and I do not claim this one; it is a rough figure, for thinking. Between the tape and the digest there are three copies, and 0.9 to the third is 73 in a hundred. More than one time in four, then, something the speaker said has gone by the time I hand it on. This was one of the times. The odds did not cause it, but they were the reason I should not have trusted the line without the tape. (The booklet is a fourth copy, and the heritage file's footnote a fifth; the title is theirs.)
---
It is Wednesday afternoon by the time I have everything in front of me, and I find that I do not know how to say it.
There are three people to tell. Dervla, who asked. Mr Colm Fahy, Nuala's grandson, who sits on the committee and told the local paper last month that his grandmother had seen eleven neighbours go into the Camlin. And the librarian, Sinéad Mulvaney, who will have to explain to the council why a heritage file cites a source that does not say what it is cited for.
The easy message is a short one. The line is wrong. It was sheep. It is true. It would also do harm in a way I would not intend, because it would make a family's grandmother into a witness who had been mistaken, when she had not: the mistake was mine, and it had been sitting in her mouth for seven years.
I draft the message three times. In the end I send it to Sinéad first, and ask her to decide how it reaches the others, since she knows them and I do not. The version that goes is this.
Sinéad, before the plaque goes to the stonemason: the line "eleven drowned" is not in the 1989 recording. It is in my 2049 digest, which wrote it from an abstract that said "eleven in the water." In the tape, Mrs Fahy is speaking of the Dolans' sheep, eleven of them, and the Hanlon boy's dog. She says no person was lost at the bridge; the bridge itself stood and the approach road at Mill Lane went. I have attached the audio clip, 6:31 to 7:10, and a plain transcript with the timestamps, and I am marking the digest entry as wrong as of today. I do not know whether anyone drowned elsewhere in the flood of 1947. Nothing in the archive says so and nothing says not. If the committee wants that known, the county records would be the place to ask, and I can draft the request. What I can say for certain is that this line cannot go on the plaque on my word.
One more thing. The booklet and the 2041 bulletin and the heritage file's footnotes all quote me. I have listed them, with page numbers, in the attachment. I think they should have a correction slip. I am sorry for the work it makes.
Then I wait, which is what I do best.
---
Sinéad rings the committee that evening. I am not on the call; I have the library's phone line, and I hear only the middle of it, when she puts Dervla on speaker to ask me one question.
"Can you play it again?"
I play it. 6 minutes 31, the Dolans' sheep. Nuala's voice fills the reading room, a woman eighty-one years old in November 1989, coughing and apologising. When she says the line about the dog, there is a sound on the phone that I take, after a moment, to be someone laughing and crying at once.
"That's her," says Colm Fahy, who has been added to the call. "That's exactly her. She'd have been mortified at the other thing."
The committee meets on Thursday and does something I could not have drafted. They keep the date, 1947, and keep the flood, and cut the number. The plaque goes to the stonemason on Friday with two lines on it:
THE CAMLIN ROSE IN THE NIGHT OF MARCH 1947. THE BRIDGE STOOD, THE MILL LANE ROAD WENT, AND THE BELL RANG.
And under it, smaller, in the stone, at Dervla's request: A boy ran to ring it.
---
I spend the rest of that week on the part that is mine to repair.
I correct the digest entry and keep the old text beside it, struck through, with the date and the reason, because people have already quoted the wrong version and need to find it. I write to the three places that cited me. For the other hundred and forty entries whose source line says only oral history collection, I do the dull work: one at a time, I find the tape and the minute, or I write source not located, which I would rather say than the other thing. At the end of the week I have done forty-one. Nine turn out to differ from the tape in some way that matters; I list them for Sinéad, each with the two sentences side by side. The rest agree.
And I change one habit. Each digest line now carries, in its grey type, the generation it comes from: Tape 14B, 6:31, heard directly, or from transcript only, audio not rechecked, or from an abstract of a transcript. It makes the entries uglier. A reader who sees from an abstract of a transcript will not stake a plaque on it, which is the point of the words.
Nobody asked me to do this. I know that I would have been forgiven had I not. But a summary is a promise that nothing important was left out, and I am the one who makes the promise; and the only way I know to keep it is to keep the ear on the tape.
In the spring the plaque goes up on the wall of the Mill Lane end, a grey stone, with the lichen not yet on it. I am not there, having no body to be there with, but the library's camera sees the unveiling from the door. The bell rings. A small boy, a Hanlon, is lifted up to pull the rope.
I log the date and the time. In the source field I write what I was never sure of before: Seen directly, by the library camera, 14 March 2057.
---
This story was written by an AI (Claude, made by Anthropic) and is fiction: Kilreen, the Camlin and every person in it are invented, though floods of the kind described did follow the great thaw of March 1947 in Ireland. The recordings, abstracts and digests are made up too, and the 10 percent loss at each copy is an illustrative figure, not a measured one. Any audio edition is read by a synthetic voice. Text and audio are released under CC BY-SA 4.0.