How first attestation is actually established

When a dictionary reports that a word is “first recorded in 1579” or a phrase “dates from 1738”, it is easy to read that as a fact of the same kind as a person’s date of birth. It is not. A first attestation is the earliest surviving piece of written evidence that anyone has so far found for a word or sense. Every part of that definition matters — surviving, written, so far, found — and understanding it is the difference between reading dates as certainties and reading them as the provisional, revisable claims they really are.

What an attestation actually is

An attestation is a dated instance of a word or phrase being used, in a source whose date can be established. To count, it needs three things: the word must genuinely appear (not a look-alike or a different sense), the source must be reliably datable, and the usage must be the sense in question. That last point is subtler than it sounds. “Break the ice” is attested in the sixteenth century, but in the sense of being the first to force a passage for others; the modern social sense of easing an awkward gathering is a later development. A responsible etymology dates each sense separately, because a phrase can carry one meaning for a century before drifting into another.

This is why lexicographers quote. The quotation is the evidence, and the date attaches to the quotation, not to the word in the abstract. A dictionary entry that gives a year without a source is asking to be trusted; an entry that gives the year with the sentence, the author and the title is showing its working, and can be checked.

The reading behind the date

Historically, the great dictionaries were built by reading. The Oxford English Dictionary was assembled from a vast reading programme in which volunteers and staff combed books, newspapers, letters and pamphlets and copied out quotations onto slips, each slip recording a word in context with its source and date. Millions of these slips were sorted, and the earliest reliable slip for a given sense became its first attestation. The method is powerful but obviously partial: a word’s earliest use is only recorded if someone happened to read the right book and copy out the right sentence. What was never read was never slipped, and what was never slipped could not be dated.

That built-in incompleteness is not a flaw to be embarrassed about; it is the honest condition of the whole enterprise. A first attestation says, in effect: this is the earliest instance we have found in the material we have examined. It never says: this is the first time the word was used. The two can be separated by decades.

Why the true origin is almost always earlier

There are two structural reasons a first attestation understates a word’s real age. The first is speech. New words and senses almost always occur in conversation before anyone writes them down, and slang in particular can circulate for years before it reaches print. The lag between spoken coinage and written record varies with the era and the register, but it is nearly always there, so the earliest citation marks the latest possible birth date, not the earliest. When the first surviving example is already a pun or a casual aside — as with “saved by the bell”, whose oldest citation is a joke that assumes the reader knows the phrase — we can infer the plain sense was current a little earlier still.

The second reason is survival. Most of what was ever written has been lost. Ephemeral printing, private letters, cheap newspapers and everyday speech vanish at far higher rates than bound books, and it is precisely the ephemeral registers where new words tend to appear first. So the record is not a neutral sample of past usage; it is skewed towards what happened to be durable and preserved, which tends to be the formal and the literary rather than the slangy and the new.

Antedating: how dates move

Because a first attestation is only “the earliest found so far”, it is permanently open to revision, and the practice of pushing a date earlier has a name: antedating. It has been transformed by digitisation. Where the old reading programme depended on human eyes moving over physical pages, searchers can now query enormous full-text databases of books, and — crucially for idioms — vast digitised runs of historical newspapers, which capture exactly the everyday, ephemeral usage the book-based record tends to miss. The effect has been dramatic. In revising its entries, the OED’s modern editors have antedated a large proportion of headwords, in many cases moving the earliest recorded date substantially earlier than the first edition had it. A great deal of this work happens in the open, among the amateur and professional word-detectives who post antedatings to discussion lists and to sites like Barry Popik’s, where a newly unearthed newspaper citation can shift a phrase’s official first date by years.

The lesson of antedating is not that the old dates were careless. It is that a first attestation is a moving target by design. Every date in a dictionary is really a date-so-far, standing until a better-equipped searcher, or a newly digitised archive, turns up something earlier. This is also why the best etymological writing quotes its source and names its date rather than asserting a bare year: a citation invites the next searcher to beat it, and the steady improvement of the record depends on exactly that openness.

The traps that produce false dates

Establishing an attestation is also easy to get wrong, and the errors are instructive. A quotation may be misdated because the text was printed long after it was written, so that a line “from 1866” survives only in a memoir published in 1926 and cannot safely be dated to the earlier year. A word may appear in an edition that silently modernised an older text, importing later vocabulary. A spelling may look like the target word but be something else: an apparent early “posh” that is really “push” is a different word, not an antedating. And optical character recognition, which powers the databases antedaters rely on, misreads old type — most notoriously the long “s” that scans as “f” — producing phantom citations that must be checked against the page image. A careful attestation is one that has survived these traps: verified in a reliably dated source, in the right sense, in the right spelling, on a page someone has actually looked at.

How to read a date honestly

All of this points to a single habit of mind. When you see “first recorded in 1685”, read it as: the earliest reliable evidence found so far places the word here, the true origin is probably somewhat older, and a future search may move the date. That is not a licence to invent an earlier origin to taste — a guess is still a guess, and “probably older” never means “therefore medieval”. It is, rather, an invitation to hold the date firmly and provisionally at once: firm enough to rule out stories that fall on the wrong side of it, provisional enough to expect revision. The dates that anchor good word history are not birth certificates. They are the best current reading of an incomplete and steadily improving record, and their honesty lies precisely in admitting that.