Making a dictionary
AMost people assume that a dictionary is a list of the words that are allowed, compiled by somebody with the authority to allow them. Lexicographers find this assumption exhausting, because their work runs in the opposite direction. The question an editor asks is not whether a word ought to exist but whether it already does, in the sense that enough people use it, in enough places, without stopping to explain what they mean. A dictionary is closer to a survey than to a rulebook.
BThe survey needs evidence, and the nature of that evidence has changed completely. Until the end of the twentieth century, most dictionaries were built from reading programmes: volunteers worked through newspapers and novels and posted in slips of paper recording interesting uses. The method produced magnificent collections and one serious distortion, because a volunteer notices a strange word and never notices an ordinary one. The slips therefore over-recorded the rare and under-recorded the common, which is precisely the wrong way round for deciding what matters.
CCorpora replaced the slips. A corpus is a very large body of real text and transcribed speech, assembled so that it can be searched by the billion words, and it answers the question the slips could not: how often, by whom, and in what company. Editors can see that a word appears mainly in sports reporting, or mainly among speakers under thirty, or almost always within three words of another particular word, and all of this goes into the entry.
DDeciding what earns a place is still a judgement. Frequency alone is not enough, because a word can be everywhere for a year and then vanish with the thing it named. The working rule in most houses is that a term should appear across different kinds of source, over several years, and be used without explanation, since an author who glosses a word is telling you it has not arrived yet.
EDefining turns out to be harder than selecting. The trap is circularity: define a chair using the word seat, and a seat using the word chair, and the reader who knew neither is no better off. Several dictionaries for learners therefore write every definition using a restricted list of about two thousand very common words, which forces editors to explain rather than to substitute, and which makes the definitions longer and much more useful.
FThen there is the order of senses, which sounds trivial and is not. A dictionary can list meanings in the order they entered the language, which tells a story, or in the order of how common they are now, which tells the reader what they probably want. Users rarely read past the first sense, so the choice decides what most people take the word to mean, and two respectable dictionaries can send a reader in different directions.
GRemoval is the one editorial act that produces letters of complaint. Print has a fixed number of pages, so something must go when something arrives, and the words that go are those the evidence no longer supports. Readers hear this as an announcement that a word has been abolished, which it is not; the word remains in older editions and in the historical record. Online editions, having no page limit, have quietly made the problem disappear, which may be the most important thing the move to screens has changed.