The Half-Life of a Mistake

D · Plate 16 in the Corrections series · Other Communications

A corrected claim does not stop existing. It persists in copies, in summaries written before the correction, in the memory of people who read the original and not the repair, and in the second-hand versions that were always going to be read more widely than either. The useful question is not whether a correction works but how quickly the incorrect version decays.

Several things determine the rate, and most of them are decided in the hours after the error is noticed rather than in the wording of the correction. This is the part of the subject that gets the least attention, because it is operational rather than editorial.

Speed matters more than completeness

An error corrected within hours has usually not been copied. The same error corrected a fortnight later exists in a dozen derivative places, none of which will be updated by the original publisher. The difference in eventual persistence between a fast rough correction and a slow polished one is very large, and it favours speed by a margin that surprises people the first time they measure it.

This argues for a low threshold to publish an interim correction. A note saying that a specific figure is incorrect and that the right one will follow shortly is a complete and useful act. It stops the claim from being repeated, which is the immediate need, and it does not require the full account to be finished first.

Correcting at the source

The corrected text should be fixed in place, marked as corrected, and dated, rather than left standing beside a correction published elsewhere. Most readers arriving at the claim after the fact arrive at the original page, not at the correction. A correction that lives only in a separate location reaches the readers who were already following closely and misses everyone else.

Where the claim has travelled, the people who repeated it should be told directly. This is unglamorous work with a measurable effect, and it is routinely skipped because it feels like asking a favour. It is not; anyone who repeated an incorrect claim in good faith has a strong interest in knowing, and being told promptly is a courtesy rather than an imposition.

What keeps a mistake alive

Three things extend the life of an incorrect claim. A correction that does not repeat the original claim, so that people who half-remember it cannot match the two. A correction that appears only in a place readers do not routinely visit. And a correction hedged enough that a reader is unsure whether the claim was wrong or merely refined.

The third is the most damaging, because it leaves the original in circulation with a small asterisk rather than removing it. Readers carrying an uncertain memory of a correction will often repeat the original claim with a qualifier attached, which is worse for everyone than either the claim or its correction alone.

The practical summary is short. Correct quickly, even partially. Correct in the place where the claim lives. Restate the claim clearly enough that someone can recognise it. Tell the people who repeated it. Nothing here is about wording, which is why the essays on wording elsewhere in this series only get you part of the way; a well-written correction that arrives late and lives somewhere else will be outlived by the error it addresses.