Dictionary wars

As I had failed in my efforts to think without recourse to language, I assumed that this was an exact equivalent of reality; I was encouraged in this misconception by the grown-ups, whom I took to be the sole depositories of absolute truth: when they defined a thing, they expressed its substance, in the sense in which one expresses the juice from a fruit

Simone de Beauvoir, Memoirs of a Dutiful Daughter

Back in July, I posted about a petition calling on Oxford University Press to change the archaic and sexist entry for ‘woman’ in one of its most widely-used dictionaries, the Oxford Dictionary (the one you get with an iPhone). A few weeks later, in August, OUP’s blog published a response to the petition from Katherine Connor Martin, whose job title is ‘head of lexical content strategy’. This response made some of the same points I had: since language reflects the culture in which it is used, a certain amount of sexism is inevitable—you can’t eliminate all bias without compromising the aim of a modern dictionary, which is to describe the way words are actually used. Martin did, however, assure readers that OUP takes their concerns very seriously, and said the dictionary’s editors were investigating ‘whether there are senses of “woman” that are not currently covered, but should be added in a future update’.

At the time this response attracted little attention. But then this week, the petition was suddenly news again. There was a spate of press reports claiming, in effect, that because of the petition, which has now attracted 30,000 signatures, OUP was going to redefine the word ‘woman’. One report, in the Bookseller, misleadingly referred to the petition as ‘an online survey’, as if Oxford had commissioned an opinion poll and was proposing to change the meaning of a word on that basis. And at that point things really kicked off: Twitter was all over it, and by Thursday the debate had even been covered by Good Morning America.

In fact, nothing Katherine Connor Martin had said the press might do–like adding new senses to the entry, or marking some synonyms for ‘woman’ more clearly as offensive–would amount to ‘redefining’ the word ‘woman’. More fundamentally, you can’t change the meaning of a word like ‘woman’ simply by altering a dictionary entry. If OUP decided to redefine ‘woman’ as ‘a mythical white equine mammal with a single horn’, it would make no difference at all to the way English speakers actually use it. It would just make the dictionary look like (if you’ll forgive the mixing of equine species) an ass. No one checks the dictionary entry for ‘woman’ before they utter a sentence containing the word: we don’t need anyone’s permission to use words to mean what we already know they mean. Dictionary-makers spend a lot of their time and energy systematically investigating what we think words mean, as evidenced by the way we use them, so they can document that fully and accurately. They take their cue from us, not vice-versa.

But most people really don’t understand that, as I found out when I waded into the debate on Twitter. The Guardian’s report had mentioned one suggestion made in the petition—though in that context it was a minor point, almost an afterthought—that the example sentences in the ‘woman’ entry might include references to different kinds of women, like ‘a trans woman, a lesbian woman, etc.’ It then quoted Katherine Connor Martin’s point about ‘senses of “woman”…that should be added in a future update’. This juxtaposition led some Guardian-reading Twitter-users to infer that OUP was planning to update the entry to include, specifically, references to trans women. And the press was then accused of caving to political pressure from the trans lobby to redefine what ‘woman’ means.

I thought that was a stretch, and I said so. Not because there hasn’t been lobbying on this issue—there has, from both sides—but I don’t think this petition is an example of it. (Its central concern is sexism rather than gender identity, and its creator is a feminist, not a trans activist.) I also don’t think that the press has ‘caved’ to any kind of pressure, except possibly the pressure to move the ‘woman’ question higher up its agenda. In the age of the internet, dictionaries, like other commercial enterprises, do feel an obligation to appear responsive to users’ concerns, so a petition with 30,000 signatures may have some influence on their order of priorities. Beyond that, I see no evidence of ‘caving’: far more has been read into Katherine Connor Martin’s comments than they actually say.

But as the discussion on Twitter progressed, I began to realise that the reason so many people believed or suspected things I found implausible or even absurd was because they saw dictionaries in a particular way—one which is at odds with the way dictionaries see themselves, and the way I, as a semi-insider, see them (I’m a linguist not a lexicographer, but I know how modern lexicography is done). Based on my conversations with people on Twitter over the last few days, some widely-held beliefs about dictionaries include the following.

The makers of dictionaries have absolute power to decide what words really mean, and they use that power selectively.

Some people I interacted with evidently imagined the editorial process at a dictionary as rather like what you see in movies or TV shows about journalism: people sitting in a meeting arguing the toss about definitions and debating whether to accept or reject new words and senses of words. (‘Does this deserve to be in the dictionary, or should is it too illogical, confusing and objectionable to make the cut?’) This may help to explain why there’s concern about the potential for political lobbying to influence definitions, just as people worry about, say, who might be putting pressure on the BBC to take a certain editorial line on Brexit.

Concern about dictionaries pushing, or being persuaded to push, a covert political agenda is reinforced by a second widely-held belief, that

The decision to include a word or sense in a dictionary is effectively an endorsement of that word or sense. It communicates that the word or sense is real, correct and should be used by everyone.

Another dictionary-related news story this week reported that ‘non-binary “they”’ (i.e., the pronoun ‘they’ used in reference to a specific individual who identifies as neither male nor female) had just been added to a leading US dictionary, Merriam-Webster. For some of my interlocutors on Twitter, this decision implied not, ‘we have looked at the evidence and found this use of “they” is now mainstream enough to be recorded in a dictionary entry describing the current usage of “they”’, but something more like, ‘we thoroughly approve of this new way of using pronouns, and we want to support non-binary people by telling everyone to adopt it’. The news reports themselves took a similar view, in some cases describing the decision as ‘validation for nonbinary people’. That only reinforced some people’s suspicion that the decision must have been prompted by lobbying.

When people expressed the view that dictionaries shouldn’t be entertaining politically partisan demands to change their entries, I pointed out that their own demand for no change was also politically partisan. But some of their replies invoked another belief which shows what’s ultimately at stake in arguments like the one about ‘woman’:

Dictionaries don’t just define words; since words (or a lot of words) name things in the world, the word-meanings that appear in dictionary entries are also definitions of reality itself.

For people who hold this belief, there is essentially no difference between asking ‘what does the word “woman” mean?’ and asking ‘what, in reality, is a woman?’ It follows that changing the definition of a word means changing what will count as reality, and in some cases that’s seen as a politically-motivated, quasi-Orwellian assault on the truth.

One of my interlocutors on Twitter explained this to me using a hypothetical question:

if some people started to demand we use the word ‘night’ to mean ‘day’ …would the dictionary record that use?

I think this person expected the answer to be ‘no, of course not, that would be absurd: “night” and “day” are antonyms in language because they’re opposites in nature’. But in fact the answer is ‘yes’: if the evidence showed that a significant section of the linguistic community was using ‘night’ to mean ‘day’, then that sense would be recorded in entries for ‘night’ (alongside the original sense, not instead of it). This is exactly what many dictionaries have done with the word ‘literally’, which, notoriously, is used very frequently as an intensifier with the opposite meaning, ‘figuratively’ or ‘metaphorically’ (as in Taylor Swift’s recent statement that ‘I’m literally breaking’, though as far as I know she remains physically intact). This ‘incorrect’ and much-criticised usage of ‘literally’ is so common, it would be hard for a dictionary whose aim is to document the actual use of words to leave it out.

For dictionaries (and linguists like me), the question ‘what does this word mean’ is an empirical question, reflecting the axiom that ‘the meaning of a word is its use in the language’. You work out the meaning (or more often, meanings) of a word by collecting and organising evidence about its use by real speakers and writers in real contexts. There are no meetings where dictionary-makers compare their intuitions about the meaning of ‘woman’, or ‘they’, and debate whether a new sense is ‘good enough’ to deserve ‘endorsement’. What they’re looking for is evidence that the new sense is sufficiently well-established to be considered part of ‘general’ or ‘common’ usage. If it is, then it belongs in the dictionary. Not because the lexicographers think it’s a worthy addition, but simply because their aim is to provide a full and accurate record of usage. If enough speakers of the language have adopted a new word or meaning, descriptive accuracy requires it to be included.

So, what is the evidence that something is part of general usage? There’s probably some variation in the criteria different dictionaries use, and also in the criteria they apply to different kinds of words, but with something like non-binary ‘they’, which has been around for quite a while in certain subcultures (e.g., on college campuses and tumblr), Merriam-Webster’s decision that the time had come to add it was likely based on evidence that it’s now turning up regularly in mainstream sources targeting a general audience, like newspapers and general interest magazines. (In fact, it has recently been adopted as a norm in the style guides of several mainstream media organisations, so its frequency in their output will probably have increased.)

The source material dictionaries treat as evidence isn’t just selected and used in an ad hoc, unsystematic way. Today’s dictionary-makers rely on corpora, which are very large (millions or even billions of words), structured, taggable and searchable samples of authentic language-use. They’re constructed to cover a wide range of written and spoken genres and media, and to represent different sections of the language-using population (for English this may include writers/speakers in all the major world regions where the language is used). From a corpus you can collect not only a large number of authentic examples of a word’s use, but also quantitative data on how frequently it’s being used in particular ways, and whether these uses are common across contexts and populations or confined to specific regions, subcultures or genres. You can also monitor changes over time, tracking the spread of an innovation or the progress of a shift in meaning.

When Katherine Connor Martin says that editors are investigating ‘whether there are senses of “woman” not currently covered [that] should be included in future updates’, I’m guessing one thing that means is that they’ll be looking at recent corpus data to see what’s changed–if certain ways of using ‘woman’ have become more salient, or if they’ve moved, like ‘they’, from the subcultural margins to the mainstream. I’m confident the evidence will confirm that ‘adult female human’ is still the primary sense of ‘woman’ (and the commonest by a long way); the question is whether there will be a case for adding anything else.

For the dictionary-makers this is a question about the linguistic facts on the ground: it’s about how people are using the word ‘woman’, how frequently it’s being used in a particular way, and what range of contexts it’s appearing in. Even if they make a show of being responsive to concerns like the ones raised in the petition, they’re not going to ignore their own data on these points: their rigorous, evidence-based approach is, after all, what enables them to market their products as ‘authoritative’. However assiduously they’re lobbied, they’re unlikely to emulate Urban Dictionary by making popular opinion the measure of a definition’s value.

But whatever the linguistic evidence shows, it won’t answer the question at the heart of the Twitter spat, which is not an empirical question about the way words are used, but an ontological question about the nature of reality. Dictionaries cannot answer that kind of question, and it’s a mistake to think they can. They can only tell us (and only claim to be able to tell us) what the word ‘woman’ is currently used to mean. That will never settle the ongoing argument about what a woman is or is not.