The Herring Trained the Horse

In 2008 Gerald Cohen and Robert Scott Ross published an account of the red herring in Comments on Etymology. The Oxford English Dictionary accepted it.

The phrase had been carrying a tidy story. A smoked herring, red from the curing, dragged across a trail to pull hounds away from a hare. The story even had a source: William Cobbett described exactly that diversion as a boyhood prank, in 1807 and again in 1833.

It made red herring feel literal. It also supplied the wrong operating model.

The earlier record is a late-seventeenth-century horsemanship text. It describes a dragged animal scent, with red herring used when the intended quarry’s scent was unavailable, to train young horses to hold a trail through the noise of an actual hunt. Nicholas Cox appears to have moved the practice from horses to hounds and inverted its purpose. Cobbett’s version made the inversion permanent.

The phrase survived. The work it referred to did not.

The literal picture had reversed the job #

That is not a small correction to an anecdote. A training aid and a decoy give opposite instructions. One asks whether an animal can hold a signal through interference. The other asks whether the signal can be broken.

The words stayed close enough to allow the substitution. The procedure was never close at all.

Institutional language creates the same exposure and creates it constantly. A risk label starts as the name of a particular instrument, booking route, entitlement or information source. The instrument is retired. The booking route moves. The entitlement is reassigned. The label stays in the report, because nothing has to happen for a phrase to stay alive.

It gets repeated by people who are not making anything up, each of whom got it from somebody using it under an earlier arrangement.

So the useful question is not whether a label has a respectable lineage. It is whether the current process still does the job the label originally named. Red herring is familiar because its received account is a story about deception. The documented case is more useful precisely because it forces a comparison of workflow: animal, signal, intended action, result.

A first attestation records use, not adoption #

Idiom reference works keep a date, and the date conceals what kind of record supplied it.

Johannes Agricola’s letter to Martin Luther in 1530 uses let the cat out of the bag. That is private correspondence. It records a phrase already natural enough to put in a letter, and says nothing about the day it entered speech.

Richard Brome’s 1652 stage play uses rain dogs and polecats. A public script asks something different of its audience: the phrase has to land when performed, in a room, at speed.

Francis Grose’s Dictionary of the Vulgar Tongue in 1785 is different again. A slang dictionary exists to record language, so its evidence bears on the compiler’s knowledge of the expression far more directly than an incidental use in a play.

Three dates, three kinds of witness, and a glossary that prints only the earliest of them has thrown away the part that says what the date means.

The distinction bites when a desk treats the first written definition of a term as its complete operating specification. An early document can establish that a phrase circulated, and in which register. It cannot establish that every later desk inherited the same inputs, controls or allocation rights.

The archive answers which record survived. It does not quietly preserve the procedure around it.

A false history can become an operating assumption #

Rule of thumb is the harder version.

James Durham’s posthumously published sermons use thumb-width as a rough measurement standard in 1685. The later account — that English common law permitted a husband to beat his wife with a stick no thicker than his thumb — has no statute and no common-law rule behind it that anyone has been able to locate.

What it has is a traceable chain. Sir Francis Buller was satirised as Judge Thumb in 1782 over an alleged remark, of which no record survives. Del Martin’s domestic-violence report put the connection into print in 1976. The US Commission on Civil Rights gave it institutional weight in 1982 by titling a report after it. Corrective scholarship followed through the 1990s.

The correction did not make the phrase less real. It separated the phrase’s documented use from an unsupported system attributed to it two centuries later.

That is the audit problem at its cleanest. A term can carry authority precisely because it sounds inherited. A model owner defends an assumption by naming it with a label whose origin sounds settled, while the mechanism being defended was either different from the start or replaced somewhere along the way without a note.

The question is never whether the label is old. It is which documented version has been selected, and what that version actually requires somebody to do.

The form outlasts the system without dating its replacement #

There is a contrary case, and it sets the limit.

Walter Scott’s The Antiquary gives the first secure attestation of cold shoulder in 1816. The meal story — an unwelcome guest served a cold shoulder of mutton — has no documentary support tying it to Scott or to any earlier custom. The literal origin is probably a bodily gesture, and the modern meaning is perfectly clear regardless.

That is not a reason to enforce a word’s oldest sense. It is the reason not to. Etymology can identify a borrowing layer or an earlier operating context. It cannot veto present usage, and it should not be asked to.

Historical linguistics has a boundary an institutional glossary does not. Regular sound change can show whether a form was inherited, borrowed early, or borrowed after a change had already run, which is why a word’s shape can date its own arrival. A name inside an organisation can be retained, redefined or revived at will, and its shape dates nothing.

So the record has to be more explicit than a linguistic one: first use, original operating definition, current implementation, deprecated senses. A report using a term without selecting among those is not merely imprecise. It makes it impossible to tell whether the analysis describes a live control or a remembered one.

A smoked herring was once dragged across a trail to teach a young horse to hold a scent through interference. Two centuries on, the same two words mean close to the opposite, and the reversal happened in print, with dates attached.

The phrase never noticed. Phrases never do.

Every institutional glossary has some of these in it, sitting in reports, being acted on. The question is which version the report in front of you is using, and whether anyone still runs that procedure.