Transcribe a Handwritten Letter for a Documentary Pitch Without Silently Repairing Its Meaning
Transcribe a Handwritten Letter for a Documentary Pitch Without Silently Repairing Its Meaning
The sentence you need is one word short, and that word is the reason you opened the file.
That is the ordinary shape of the problem. A letter reads cleanly for eight lines, and then a name goes soft under a fold, or a stain eats three letters, and the part your pitch depends on is exactly that name. The page is not going to give it to you. What the page can give you is a transcription that shows the gap and says so — and that is worth more to a documentary pitch than a smooth sentence with an invented noun sitting in the middle of it.
A transcription is not a tidy-up. It is a report on what is on the paper, including what cannot be read. The National Archives' Citizen Archivist transcription guidance is built on a version of this discipline: type what you see, keep your own comments separate from the document's words, compare letters within the same hand, and mark a word you cannot read rather than supplying one. Those are the conventions of that particular catalog, not a universal standard, and nothing in them settles your rights to quote anything. But the underlying habit travels well, and it is the habit this article is about.
So let me give you something concrete to work on. The page below is invented. I made it up for this article: a single sheet of lined paper, written on one side in brown ink by a hurried hand, addressed "Dear Sister Nell," signed "E. Bourke." It carries a struck-through phrase, a word added above the line, and one name that cannot be read. No such document exists, and there is no image of it anywhere. Everything that follows is a worked example of method, not a report of a real transcription.
Keep the letter visible and identifiable
Before you type a single word, write down what you are looking at. Which repository, which collection, which folder, which item number. Which leaf, if the letter has more than one. Recto or verso — front or back. Who granted the permission, and for what use. That header is not bureaucracy; it is the thing that lets a producer, a fact-checker, or you in eight months find the page again and see whether you read it the same way.
Work beside the full page, not only beside a crop. A crop is a legitimate tool — enlarging a single word is often the only way to see whether a faint stroke is a pen line or a crease — but a crop also removes the sentence around the word, and the sentence is where half your evidence lives. A crop cannot tell you that the line above continues, or that what looked like a full stop is the top of a letter from the line beneath. It also tempts you to imagine a stroke that isn't there, because enlarged, every fibre of the paper starts to look like ink.
Keep the supplied image unchanged. Do not save an enhanced or thresholded copy over the original, and do not let an enhanced copy become your reference. Contrast and sharpening can invent edges and erase faint ones. If you make a working copy for enlargement, label it as a working copy and go back to the original to check every hard call.
And keep permission separate from legibility. Being allowed to photograph a letter does not mean you are allowed to reproduce it in a pitch deck, and being unable to read a letter does not make it any less protected. Those are two different questions, asked of two different people.
Read the hand before completing the sentence
Start where the page is clear. A writer's hand is consistent, so the clear stretches become your dictionary: this is how she forms an n, this is how she closes an o, this is what her w looks like when she is hurrying. Then, when you hit something doubtful, you compare it to the clear instances rather than to your expectations.
On the fictional sheet, the small word sitting above the line between "was" and "work" is the first test. It is written small and fast, and in isolation it could be several things. But its two letter-shapes match the n of "noon" and the o of "town," both of which she wrote clearly twenty words earlier. The comparison does not make the word certain; it makes it supported. There is a difference, and you should keep it.
Now the harder discipline. On the same sheet, in the sentence "Mr. ______ came out with a paper and read it twice," the name is unreadable: five or six letter-shapes covered by pooled ink at a fold, with the first shape too damaged to classify as upper or lower case. Context can tell you some things. Someone came out with a paper, and that suggests an official, an owner, a manager — a person with standing to read something aloud. Context narrows the role. It does not spell the name, and no amount of staring at a role will produce letters that are not on the page.
This is where the temptation gets specific and dangerous. Suppose your pitch background notes already name a mill agent, Hollis, active at that place and time. Hollis is six letters. Hollis fits the gap. The sentence suddenly reads beautifully. That is precisely the moment to stop, because you have stopped transcribing and started remembering.
Do not let an expected name, a plausible date, or a familiar phrase supply a missing stroke. When the visible evidence cannot distinguish between candidates, say so, and say why: the ink has pooled, the fold crosses the letters, the initial is unresolvable. An uncertainty you can explain is usable. A certainty you cannot demonstrate is a liability in a film.
Declare a small, consistent notation
Pick a short set of marks and write the key at the top of the file, so that anyone opening the transcription knows what your brackets mean without asking you. Here is one that is small enough to keep straight and clear enough to survive a handoff:
[del: ...] text struck through in the document, given as written
[ins: ...] text added above or below the line
[illeg.] one word that cannot be read
[?word] a reading the marks support but do not settle
The full literal transcription of the invented leaf, under that key, looks like this:
Dear Sister Nell,
I write in a hurry for the post goes at four. On Tuesday the gates were
shut at noon and [del: I think] I am sure the whole town was there by
dark. There was [ins: no] work for any of us after that. Mr. [illeg.]
came out with a paper and read it twice. Kate says the word was wages
and that we are not to be taken back.
your loving sister
E. Bourke
Three things are load-bearing here, and all three are easy to lose.
The deletion keeps both versions. She wrote "I think," struck a single line through it, and wrote "I am sure" after. If you transcribe only "I am sure," you have erased a revision the page actually records — and the fact that this writer first hedged and then strengthened her claim is evidence about the document, even if interpreting it is a separate job. Present the wording; leave the psychology to the argument.
The insertion is marked even though you could quietly resolve it. Without it the transcript would read "there was work for any of us after that," which reverses the meaning. But marking it also records that the "no" was added after the first draft of the sentence, which is a fact about the page.
The unreadable word stays unreadable. Not [?Hollis], which would claim the marks point somewhere, and not [illeg.] quietly upgraded to a name because the paragraph reads better. The gap is the finding.
Two more rules keep this honest. Preserve visible spelling and punctuation; do not modernize, and do not add apostrophes that the writer left out. And keep your commentary out of the transcription. The line "this word is probably a surname" belongs in a separate note file, not in the middle of the source text.
Notation collides with real documents more often than people expect. If the letter itself contains square brackets or parentheses, transcribe them as written, and switch your editorial marks to something else so that a reader cannot confuse the writer's punctuation with your key. And name your convention rather than implying it is the standard: the National Archives Catalog uses its own marks for its own searchability, and a scholarly edition uses another set again. Yours needs to be declared, not assumed.
Reread the consequential words against the image
When the transcript is done, read the finished passage back against the page — and read it looking for the words that would change your pitch if you had them wrong.
Negatives first. In the sentence "we are not to be taken back," the "not" is small and compressed, and it is the single most consequential mark on the sheet: it means the workers are shut out, not recalled. Drop it and the pitch's entire premise inverts. Read that word twice, then have someone else read it.
Names and dates next. The name is unresolvable, and — check this carefully — there is no date anywhere on the leaf. Not a head, not a tail, not a "March 12th" tucked above the salutation. If your pitch says "that March," that date came from somewhere else, and you should be able to point at where.
Then get a second reader, and get them honestly: show them the image without telling them the candidate first. If you say "I think this is Hollis," you will get agreement, and the agreement will be worthless. Agreement is useful when it is independent. Even then, two readers landing on the same guess has not turned an ambiguous mark into certainty — it has produced two guesses. Keep the alternative open when it materially changes the meaning.
Move from transcription to a pitch quotation without erasing the work
You now have two documents, not one. There is the literal transcription, with its marks and its gap. And there can be a normalized reading copy, kept separate and labeled as a convenience — resolving the strike and the insertion, lightening the punctuation, and still leaving the gap visible:
Reading copy (normalized; not the transcription)
Dear Sister Nell,
I write in a hurry for the post goes at four. On Tuesday the gates were
shut at noon, and I am sure the whole town was there by dark. There was
no work for any of us after that. Mr. [unreadable] came out with a paper
and read it twice. Kate says the word was wages and that we are not to be
taken back.
your loving sister
E. Bourke
The reading copy changes no word and fills no gap. It exists so a colleague can follow the sense, not so the pitch can quote a smoother text than the page supports.
From there, the pitch quotation has to point back to the stated witness and expose its own additions. A defensible version:
The letter reports that the gates "were shut at noon" and that "there was no work for any of us after that." A man the writer does not name "came out with a paper and read it twice." The account the writer relays from Kate: "we are not to be taken back."
Now the polished version, which is the one to catch yourself writing:
On 12 March, the mill agent, Mr. Hollis, read the wage notice aloud to the hands, and the letter closes: "we are not to be taken back."
Everything wrong here is invisible. The date does not appear on the page. The name does not appear on the page. "The hands" is not what she wrote — she wrote that the whole town was there. Kate's relayed report has become narration. And the quotation marks now sit around words that appear on the page only as Kate's relayed report, presented as though the letter closed with them — so no reader of the pitch can tell where the evidence stopped and the drafting began.
Withhold the claim that depends on the unreadable word. If the pitch needs a named man reading a notice, that claim has to rest on something else — a newspaper, a company record, a second letter — and until it does, the pitch says "a man she does not name."
One last boundary, and it is the one that gets forgotten when a transcript finally looks clean. A good transcription establishes what the document appears to say. It does not establish that the writer's account is true, and it does not authenticate the signature at the bottom. "E. Bourke" written on a page is a name written on a page. Whether E. Bourke wrote it, whether E. Bourke was there, and whether the gates were in fact shut at noon are three further questions, each with its own evidence. The transcriber's job ends before all three.
What goes in the research packet: the image with its identifier and permission note; the literal transcription; the notation key; the note file; the reading copy, clearly labeled; and the pitch sentence, with its gap showing.
And one pitch-use decision, made: quote the two sentences that rest on legible marks — the shut gates, the absence of work, the relayed "not to be taken back." Describe the notice-reading without naming the man, and move the identification of Hollis into the research questions, where it can be chased against records that might actually spell it.
The gap in the middle of that paragraph is not a failure of the transcription. It is the most accurate thing in it. Leave it where a reader can see it, and the rest of the page will hold.
Frequently asked questions
What should be recorded before transcribing a handwritten letter?
Write down the repository, collection, folder, and item number; which leaf, if there is more than one; recto or verso; who granted permission and for what use. Work beside the full page, not only a crop, because a crop removes the sentence around a word. Keep the supplied image unchanged, do not save an enhanced or thresholded copy over the original, and do not let an enhanced copy become your reference. Keep permission separate from legibility.
How should I handle an unreadable name when context suggests who it might be?
Context can narrow the role, but it does not spell the name. Do not let an expected name, a plausible date, or a familiar phrase supply a missing stroke. When visible evidence cannot distinguish between candidates, say so and say why: the ink has pooled, the fold crosses the letters, the initial is unresolvable. Under the article's notation, [illeg.] marks one word that cannot be read, while [?word] marks a reading the marks support but do not settle.
Why should a transcription preserve deletion and insertion marks?
The deletion keeps both versions. If you transcribe only 'I am sure,' you erase a revision the page actually records: the writer first hedged with 'I think' and then strengthened the claim. The insertion is marked even though it could be quietly resolved; without it the transcript would read 'there was work for any of us after that,' reversing the meaning. Marking it also records that 'no' was added after the first draft. Preserve visible spelling and punctuation, and keep commentary out of the transcription.
How should a pitch quotation relate to the literal transcription?
You can keep two documents: the literal transcription with its marks and gap, and a normalized reading copy kept separate and labeled as a convenience. The reading copy changes no word and fills no gap. A pitch quotation has to point back to the stated witness and expose its own additions. A defensible version quotes the shut gates, the absence of work, the unnamed man who 'came out with a paper and read it twice,' and Kate's relayed 'we are not to be taken back.' A polished version that invents a date, names Hollis, says 'the hands,' and presents Kate's relayed report as narration and quotation as the letter's close is misleading.
Does a good transcription prove the letter's account is true or authenticate the signature?
No. A transcription establishes what the document appears to say. It does not establish that the writer's account is true, and it does not authenticate the signature at the bottom. 'E. Bourke' written on a page is a name written on a page. Whether E. Bourke wrote it, whether E. Bourke was there, and whether the gates were in fact shut at noon are further questions, each with its own evidence.