Primary, Secondary and the Gap

A primary source is not a truthful one. This is the first thing to get straight, because the word sounds like a quality rating and is a statement about position: a primary source was produced within the period being studied, by someone in a position to know something about it. A tax register is primary. A propaganda poster is primary. A diary kept by a man determined to make himself look good is primary, and so is the letter in which he lies to his brother about the same week.

What historians do with such things is not believe them. It is ask what each one was for.

Consider two documents about the same medieval village. The first is a chronicle written by a monk, describing the piety of the local lord. The second is a manorial court roll listing fines for letting pigs into the wrong field. The chronicle is trying to persuade you of something, and its author knew that his abbey's relationship with that lord depended on how the passage read. The court roll is trying to get through Thursday.

Both are biased. The difference is the shape of the bias. The chronicle's distortions point in a direction and were chosen; the court roll's are structural. It records what the manorial court had jurisdiction over and nothing else, it under-records anyone the court could not reach, and it flattens every dispute into the categories the clerk had available. Structural bias is correctable in a way that rhetorical bias is not, because you can work out what the source was built to miss.

This is why the single most useful move in the discipline has a name: reading against the grain. You read a source not for what its author set out to establish but for what they took so entirely for granted that they did not think to argue for it. A court record that mentions, in passing, that a woman was running the workshop when the accident happened is not making a claim about women's work. It is assuming it. That incidental detail is stronger evidence than a contemporary treatise insisting women never worked, because nobody was trying to make you believe it.

Almost everything social history has recovered in the last sixty years came out of sources read this way: people who left no writing of their own, appearing in the margins of documents made about them by institutions that wanted something else.

The obvious objection is that this licenses reading anything into anything. It would, without the second rule, which is corroboration. A single vivid first-hand account is exactly as unverifiable as a thin one and considerably more persuasive, which is the dangerous combination: detail feels like proof and is not. What converts a source into evidence is a second source, produced for a different reason by someone with different interests, that happens to agree. One document is an anecdote, however good it is.

And then there is the chain of transmission: the sequence from the event to the text in front of you. A speech is given. Somebody takes notes, or does not. The notes are written up, or reconstructed from memory. A version is printed, decades later, by an editor with a view. Each link is a place where the words could change, and a great many famous quotations turn out to have a forty-year gap somewhere in that chain. The event may well have happened. The wording is a separate question, and it is the one that almost nobody asks.