Field guide to quote context and sources collection: the fields

Findings evaporate. Somebody spends two hours establishing that a line has no printed history before a particular decade, tells three colleagues, and eighteen months later the same line arrives from a different direction and the work is done again from scratch. The fix is not more research. It is a record with fields in it.

This page describes what a quotation record should carry, why each field earns its place, and where files of this kind usually go wrong.

What to take away

The fields

Canonical string. One wording, chosen as the entry's headword, written exactly. This is a filing decision, not a claim about correctness.

Variants. Every other wording you have seen, each with where you saw it and its date. This is the field people skip and the one that does the most work later, because the pattern of variation is itself evidence: a line that varies freely was transmitted orally or by retelling, and a line that is stable came off a page.

Earliest located appearance. A date, a source, and what that source attaches to the line. Not the earliest famous appearance. The earliest one you found, however obscure.

Attribution history. Who the line has been credited to, and from when. Kept as a sequence rather than as a single name, because the sequence is what shows credit migrating toward better known figures.

Primary source status. Located, not located, or not applicable. Where located, an edition and a locator that someone else can open.

Grade. Your confidence, on whatever scale you use. A workable scale runs A to D: A for a located primary source, B for strong secondary support with no primary source, C for a single weak witness, D for a name with nothing behind it. Our grading of evidence rather than verdicts sets out one.

Search record. Which strings, which corpora, which date ranges, and when you searched. Name them: Google Books, HathiTrust, the Internet Archive, Chronicling America, Project Gutenberg. This is the field that makes a null result reusable.

Rights note. Separate from everything above. Whether you may reproduce the passage is a legal question, not a scholarly one, and it turns on facts about the specific work; the copyright office's account of how long protection lasts is the place to start rather than a rule of thumb.

Decision. What you did: used with a label, used without a name, or dropped. Recording the decision prevents the same argument being reopened by whoever forgets it was settled.

One record filled in shows the fields at working length.

Why the search field matters most

Every other field describes the world, and the world does not change much. The search field describes your coverage, and coverage improves constantly as archives digitize.

A record that says "not found" is dead weight in three years. A record that says which strings were tested against which corpora, on a stated date, tells a later reader exactly what to redo and what to skip. That is the difference between a file that appreciates and one that rots.

The account of how the digitized record thins out by period is the reference for judging which of your old searches are now worth repeating. The coverage statements published alongside collections such as the Library of Congress newspaper research guide are the model for writing your own.

Shape of the record

Two structures work and one does not.

The third is what most people build, because it is the obvious shape and a spreadsheet invites it. It destroys the evidence on every update, so the file can never be audited and a disputed finding cannot be reconstructed.

If a spreadsheet is what you have, the workable compromise is one sheet of lines and one sheet of appearances, joined on an identifier. The habit of never overwriting matters more than the technology.

Where these files go wrong

Using the file honestly

A private research file is not a reference work, and the transition between those is where trouble starts. Once findings are published, readers will treat every field as an assertion, including the ones you meant as working notes.

A grade plus its search record is a claim a reader can weigh. A grade alone is an authority claim, the same one that produced the problem you are fixing.

The labels that distinguish a sourced entry from an attributed one belong in the published view and the internal one.

Where the eventual answer is that a line is a common saying rather than anyone's sentence, record it that way rather than leaving the attribution field empty. Our material on sayings that have histories instead of authors covers what those entries should look like.

Common questions

Is this not excessive for a handful of lines?

The fields cost about two minutes each to fill and save hours on the second encounter. For a handful of lines you will use once, skip it. For anything you publish, the search record alone justifies the file.

What identifier should a line have?

Anything stable that is not the wording, since the wording is the thing that varies. A short arbitrary code works and avoids the trap of filing by a version you later decide is wrong.

Should the file record who did the check?

Yes, and when. Not for blame: for follow-up. The person who searched knows things that did not make it into the fields.

How should disagreements between colleagues be stored?

As two findings with their apparatus, not as one resolved answer. Disagreement between careful people is usually a signal that the evidence is genuinely thin, and flattening it hides that.