Unnamed sources

Why does a reporter's promise never to name a source yield more truth rather than less?

There is something backwards about anonymous sourcing. A reader is asked to believe a claim whose author has been deliberately hidden, and the hiding is done by the very person asking for the belief. Every instinct says this should lower the truth content of the news: remove the name, remove the accountability, and you have built a channel for rumour.

And yet the reporting that has mattered most — the memo nobody was supposed to see, the safety report that was buried, the count of bodies that did not match the official count — has usually arrived through exactly that channel. So the question is not whether anonymity is risky. It plainly is. The question is what it is doing that makes it, on balance, produce more true things than it suppresses.

Start with the source rather than the reporter. Who has the information you would most like to read? Almost always someone inside: an auditor, a nurse, a mid-level official, a contractor. Now ask what happens to that person the day after they hand it over under their own name. They are identifiable to the one institution best placed to punish them, and the punishment does not need to be dramatic to work — reassignment, a lapsed contract, a reference that never arrives.

So notice the shape of the trade. The value of the information to the public rises with how badly the institution wants it hidden. But the cost to the source rises with the same quantity. The more worth telling a thing is, the more it costs to tell it. If that cost falls entirely on the teller, the highest-value information is exactly the information that never moves.

Then what would move it? Only something that lowers the cost. Anonymity is that something — but here is the step that matters: a promise only lowers the cost if the source believes it will hold when holding it is expensive. Consider a reporter who promises protection but adds, quietly, "unless a court insists." The source now has to price in the court. That is not a discount; that is the same risk wearing a coat. The promise has value in proportion to how unconditional it is.

That is why the practice took the shape it did. In the United States there is no general federal reporter's privilege — Branzburg v. Hayes (1972) declined to find one in the First Amendment — and while nearly every state has a shield law or equivalent court-made protection, federal proceedings have repeatedly left journalists exposed. The gap has been filled not by law but by demonstrated willingness to absorb the penalty: Judith Miller spent 85 days in jail in 2005 rather than identify a source. You can dislike the particular case and still see what it does to the general market. Each such episode is a costly signal, and its audience is not the court but the next source deciding whether the promise is real.

Which raises a question worth sitting with: if the promise is unconditional, what stops it from being abused? Nothing, structurally. That is the honest cost. An anonymous quote is a way for a powerful official to float a claim and keep the deniability — and much of the anonymous sourcing that fills political coverage is exactly this, not whistleblowing. The commitment that unlocks the nurse also shelters the spin doctor.

So the practice has to be paired with something that separates the two, and it is: the editorial layer. Serious outlets require the identity to be known to the editor, require the reason for anonymity to be stated to the reader, and require corroboration from documents or a second source before the claim runs. The promise is unconditional as to the public and to the state; it is not unconditional inside the newsroom. That asymmetry is the whole design.

The tempting misreading is that anonymity is a favour granted to a source. It is better read as a commitment made by the reporter, and commitments are only worth what they cost to break. This is why the promise is treated as near-absolute in newsroom codes while so much else is negotiable, and why breaking one is a career-ending act rather than an editorial judgement call.

Two refinements. First, "anonymous" is not one thing. On the record, on background (usable, attributed to a described role), deep background (usable, unattributed), and off the record (not usable at all) are different contracts, and the drift between them is where most disputes actually happen — so the terms are supposed to be agreed before the information is spoken, not after.

Second, anonymity does not make a claim true. It removes one of the reader's best filters and obliges the outlet to replace it. When you see "according to three people familiar with the matter," the number is not decoration; it is the substitute filter being shown to you. Whether the substitute is adequate is genuinely contested, and the profession's own critics argue it is granted far too routinely for claims that carry no risk at all to the speaker.

• Shield laws: why legal protection varies so sharply by jurisdiction, and whether statutory protection weakens the costly-signal mechanism by making the promise cheap.
• The abuse case: how outlets' stated anonymity policies differ, and whether "familiar with the matter" sourcing has measurably degraded.
• Leak-hunting through metadata rather than testimony — what a promise can still protect when the trail is technical.

Every term the collection defines is gathered in the glossary.