I first learned about this phenomenon when I was eighteen, in a linguistics class that I took about Canadian English in Kingston, Ontario. It stood out for me among the sea of dialect maps and survey methodologies because I realized that I’d done this exact thing. I was a child who sang “zee” at the end of the alphabet song in the 1990s, until some point around middle school when I switched to using “zed” consistently. What’s more, I was still slightly embarrassed by this fact and had done my best to put it out of my mind, because clearly I should never have been using the un-Canadian “zee” in the first place. When I realized that I’d done this, I asked my mom what she called the last letter of the alphabet. I’ve only ever known her as a zed-sayer, but apparently she’d done the same shift long before I was born. My switch from “zee” to “zed” happened at about the same age that I started consistently using Canadian spellings in words like “centre” and “colour” instead of “center” and “color.” I don’t recall anyone telling me to do it, but I do recall it being a conscious choice, fueled by that exact sense of social identity that Chambers described. At the time, acquiring a sense of linguistic nationalism was a way of going with the flow, of following the dominant usage of my parents and teachers. In adulthood, especially on the internet, I use Canadian spellings in my posts and messages partly out of habit, but partly also because it goes against the flow: it’s a subtle way of reclaiming space against the idea that all English speakers on the internet fit neatly into the choice I’ve faced in so many dropdown menus between “American” and “British.”
We all make linguistic decisions like this all the time. Sometimes, we decide to align ourselves with the existing holders of power by talking like they do, so we can seem rich or educated or upwardly mobile. Sometimes, we decide to align ourselves with particular less powerful groups, to show that we belong and to seem cool, antiauthoritarian, or not stuck-up.
The most legendary study of social factors in language differences is about how much people of different social classes use the stereotypical “New Yawk” accent, with the R dropped from after the vowel. In November 1962, linguist William Labov went into various department stores in New York City and asked how to find something—the shoe section, for example—that he already knew was on the fourth floor. The salesperson would reply “fourth floor” or “fawth flaw” and then Labov would pretend not to have heard, getting the salesperson to repeat the location more carefully. After this exchange, Labov would head off in the appropriate direction, but not to buy shoes. As soon as he was out of sight, he’d pull out a notebook and record whether the salesperson pronounced the R in “fourth” and “floor.” He found that, sure enough, the salespeople at the fanciest department store, Saks Fifth Avenue, said R more than those at the mid-range one (Macy’s), who in turn used R more than the bargain store (the now defunct Klein’s), and that people also tended to pronounce R more in careful speech, when he asked them to repeat themselves, than they had the first time around. But it’s hard to shop at Saks Fifth Avenue on a retail salary: the salespeople themselves came from similar class backgrounds across all three stores. Instead, it was their perception of the kind of customer they catered to that made the difference, even though Labov took pains to report that he dressed the same in all places: “in middle-class style, with jacket, white shirt and tie, and used my normal pronunciation as a college-educated native of New Jersey (r-pronouncing).” (One assumes that the pronunciations might have varied even further if he’d dressed up or dressed down.)
But where did our ideas of what sounds upper or lower class even come from? In New York City, the R-less pronunciation is less prestigious. Although it’s a feature of many American varieties, such as Boston English, African American English, and Southern American English, it’s not favored in media. When people in the United States talk about “losing an accent,” they often mean gaining an R in words like “fourth floor.”
If we hopped across the Atlantic and did the same study in British department stores, however—if we went to, say, Harrods and Debenhams and Poundland—we could find the inverse. Salespeople at Harrods, the poshest of the posh, would have no Rs at all, whereas staff at Poundland, where (almost) everything costs a pound, might have Rs if we picked our city carefully, such as Bristol or Southampton. R-ful varieties are found in parts of Britain, including Scotland and Northern England, but they’re not favored in London or on the BBC. English speakers don’t all talk like our books and media any more than actual French and German speakers talk like the model dialogues in language-learning textbooks. When people in the UK talk about “losing an accent,” they often mean losing the R in words like “fourth floor.”
Clearly, it’s not R’s fault. R is a harmless consonant that never asked to be embroiled in any of our petty human squabbles. Rather, it’s what we take R to mean in different contexts. It’s like how blue can signal a sports team, a cold-water setting, a hyperlink, a period in the life of Picasso, and so on. R in itself is neither good nor bad: its meaning, and the meaning of the accents that do or do not have it, is constructed by society. Like how money is just squiggles on paper or on a screen until it determines whether you can eat lunch, words are just meat twitches until they determine whether you can get a job—or whether someone will even deign to tell you where the shoe section is. If we all woke up tomorrow and decided that every single vowel sounded better with an R after it, we could make it happen. (Ermargerd, whart ar world thart wourld ber.)
But we don’t generally wake up and decide to change our minds about R. Instead, we get our social linguistic cues from the people and power dynamics around us. One vivid example of this power dynamic comes via James Milroy, who we last saw comparing social networks in England and Iceland. In the story of a language, just like everywhere else, history is written by the winners: Milroy recounts a typical attitude from an influential historian of English named H. C. Wyld in 1927, who “was quite insistent that the only worthy object of our study was Received Standard English. . . . The language of ‘the Oxford Common Room and the Officers’ mess’ is an appropriate object of study, whereas that of ‘illiterate peasants’ is not.” You practically want to reach back through time and punch the elitism.
Wyld wasn’t the first linguistic elitist: before there was the elite Oxford Common Room, there was the Roman forum. The Romans, good at roads and aqueducts and armies, also left a legacy of writing: for over a millennium after the fall of the Roman Empire, if you were educated, you learned Latin. To be an English writer in the era when formal writing was shifting over from Latin to English was to be a self-hating English writer: anything you could do to make English more Latin-like would also make it better. Robert Lowth, who wrote a widely used English grammar in 1762, culled examples of so-called false syntax from luminaries of English writing like Shakespeare, Milton, and the King James Bible—not as hints that perhaps English grammar was actually just fine as it was, but as cautionary tales about how even the greats should have been more Latin-y.
It was like a competition to see who could be the most uptight. Lowth gave us an early suggestion against the sentence-ending preposition: “This is an Idiom which our language is strongly inclined to; it prevails in common conversation, and suits very well with the familiar style in writing; but the placing of the Preposition before the Relative is more graceful, as well as more perspicuous; and agrees much better with the solemn and elevated Style.” Lowth himself wasn’t completely against it (after all, he used it himself in “strongly inclined to”), just passing an aesthetic judgment. But later grammarians elevated this preference into a full-on ban, and by similarly specious reasoning objected to infinitive splitting and “they” as singular, despite centuries of prior English usage. The same Latin-worshipping tradition was responsible for adding superfluous silent letters to words like “dete,” “samoun,” and “iland,” because “debt,” “salmon,” and “island” look more like Latin “debitum,” “salmonem,” and “insula.” Ne
ver mind that “island” doesn’t even come from Latin, or that generations of schoolchildren would now have to go to extra effort. Many languages can’t have spelling bees because their spelling systems are so logical that no one would ever get knocked out. English spellers can only dream!
We could almost feel sorry for the depths of self-loathing that these grammarians must have felt, to be so determined to replace their own language’s forms with that of another, if it weren’t for how they infected us with it as well. While they didn’t wholly succeed at the grammatical side, especially in speech and among skilled writers who trusted their own ear or felt they knew enough to break the rules, they did leave us with a vague sense of unease at the whole prospect of the written word. Even after years of writing, most of us have a hard time trusting what we naturally think sounds like a reasonable English sentence, haunted as we are by the ghosts of misguided grammarians.
But while modern linguistics has moved on, and even modern writing manuals are scraping off the heavy lacquer of Latinization with more or less enthusiasm, we’ve acquired a new form of linguistic authority on our digital devices. Tools like spellcheck, grammarcheck, autocomplete, and speech-to-text impose someone’s ideas of the rules of English automatically—invisible authorities that we can defy but not avoid. If a writing handbook like Lowth’s or Strunk and White’s displeases you, you can throw it across the room or leave it to gather dust, but when you want to type a word that’s not in a predictive text model, you’ll fight for every letter. In her book Fixing English, Anne Curzan describes how Microsoft Word’s grammarcheck continues to perpetuate this same kind of discredited, Latin-based style advice and how her colleagues in the English department, while considering themselves sufficiently expert in writing to ignore or turn off the green squiggles, had still never wondered where the grammar advice came from. If English professors who question the authority of texts for a living haven’t thought to question the origins of their invisible electronic grammarians, what possible hope do the rest of us have?
Language features are not neutral in the way that the calculator feature is neutral. “Standard” language and “correct” spelling are collective agreements, not eternal truths, and collective agreements can change. Communication tools that expose us to more people may speed up the spread of new words, but tools that aim to help us with language can also slow down natural linguistic evolution by nudging us towards the versions that have already been programmed into the device.
I’m convinced that spellcheck is responsible for people’s consistent misspelling of my surname: my spelling, “McCulloch,” is never found in spellcheck by default, but the very similar name “McCullough” is always there instead, and when people misspell my name on a computer, they always pick the spellcheck version. Conversely, people occasionally misspell my first name, Gretchen, when writing by hand, but never do so when spellcheck is available. It seems that my names belong to two different classes of digital citizenship: one supported by the machine and the other rejected by it. This might seem relatively harmless given my German first name and Scottish surname, but I expect that if we looked at which names are found in autocorrect and autocomplete, we’d find that typical English names would be well represented and names from other languages less so. At a societal level, it’s a case of bias-laundering through technology that serves to reinforce people and names that are already powerful.
Default computer spellings are powerful enough to have created a shift in British English since the 1990s: while American English prefers a Z in words like “organize” and “realize,” British English has traditionally used both -ise and -ize spellings. But spellchecks have tried to prevent people from spelling the same word differently within the same document by enforcing “organise” and “realise” all the time when set to British English, leading to an upswing in -ise endings among the general British typing public and the perception that -ize is only for Americans.
In writing this book, I’m therefore very aware that upholding the old-school Latin worship is a political decision, just as it would be if I decided to go full-on grammar anarchist. I think it’s important to be upfront about such things, especially in an age when everything from books to tweets may later be mined to prove how common or acceptable a particular usage was at a particular time. Yes, I’m writing for you, the reader, but in another sense we’re all writing for the unblinking eye of Data. If the most enduring legacy of this book is the slight shifting of a point on a line graph in some yet unborn person’s analysis of this decade in the English language, I want to be deliberate about which direction I’m shifting that point in. What I’ve seen from several editors and lexicographers is the realization that we’re becoming trapped in a loop: dictionaries and writing manuals refer to edited prose in order to determine what is “standard” English, but the creators of such prose refer back to the same dictionaries and manuals in their editing, each waiting for the other to move first. I’ve decided to play my part in correcting for this bias by opting for the more innovative direction wherever I perceive a choice: going towards where I think edited English prose will be by the end of the century, catering to the reader of the future rather than the reader of the past. As a reader and analyst of data myself, I get a joyful thrill every time I zoom out on the English language and realize that we’re somewhere in the middle of its story, not at the beginning or end. I don’t know how we’ll be writing in the twenty-second century, but I feel a responsibility to help its linguists gain a broad cross-section of the language of the twenty-first by not lingering overlong in the twentieth.
To that end, I’ve chosen to lowercase “internet” and social acronyms like “lol” and “omg” and to write “email” rather than “e-mail,” and when I’ve needed to make a decision on other spelling choices, I’ve looked up which ones are more common in the Corpus of Global Web-Based English and tweets by ordinary people rather than which ones are favored by usage manuals, which has led me to close many compound words. (While I was working on this book, the Associated Press switched its recommendation from “Internet” to “internet,” so I have every expectation that any similar judgment call I make will seem boring within a decade.) I’ve adopted the retronym “networked computers” for what were formerly called small-i internets, and I talk about “websites” rather than trying to insist on a distinction between “the internet” and “the worldwide web” which is no longer active for younger and nontechnical users. (I avoid the now dated-sounding “the Web” or “the Net” entirely, and reserve “cyberspace” for jocular historical use.) I’ve also included a substantial proportion of absolute time references rather than relative ones, aiming to be precise about whether I think something is true of the early twenty-first century, the 2010s, a specific year, and so on, rather than saying “now” or “currently” and requiring readers to flip to the copyright page and subtract a year or two for preparation, as I’ve had to do many times when reading other sources. I’ve freely used the singular “they,” and split what infinitives needed splitting, and preserved all spelling and typographical choices found in quotes from other people, but I’ve otherwise kept to standard bookish spelling and capitalization and punctuation, and even suffered to have my Canadian spellings changed for US audiences. But, although it’s common internet usage, I have not lowercased names of internet companies and platforms like Facebook and Twitter and YouTube.
Despite my many objections, I still use spellcheck and predictive text. Most of the time, they’re pretty useful! I don’t have to remember the c-to-s ratio in “necessary” or the exceptions to the “i before e” rule, which surely frees up valuable brain cells, and I can simply add words like lowercase “internet” to my phone’s dictionary. But I also wonder what a world would look like where none of us cared about such things in the first place. From a linguistic perspective, all varieties are equally worthy: every language and dialect is just as much a manifestation of the incredible human language ability that
is our birthright as a species. You wouldn’t say that some birds aren’t singing right just because they’re lower in the (ahem) pecking order. No more are certain ways of speaking inherently inferior. Could we not put our tremendous computing power (both human and mechanical) to better use than upholding the prejudices of a bunch of aristocrats from the eighteenth century?
Some technolinguistic tools have been attempting to do just that, albeit with mixed results. Wikipedia, whose slogan is “the free encyclopedia that anyone can edit,” has been very effective at combating obvious vandalism with dedicated volunteer editors, but faces more subtle problems of bias in what it covers, because the volunteer editors it attracts are disproportionately male, well-off, and English speaking, and they tend to edit topics they’re already interested in. Google Docs, where this book was written, has a spellcheck that draws on internet data, sometimes with surprising results. Once, to my great joy, it proffered a more common spelling of “Ronbledore” (an obscure Harry Potter fan theory that Ron Weasley is actually a time-traveling Dumbledore). Other times, it has persisted in suggesting the closed spelling “alot” over the open spelling “a lot”—a version that’s common but more informal than I’d expect a spellcheck system to endorse. Perhaps the most promising computational tool for fighting bias rather than reinforcing it is Textio. This is a startup that assesses the text of your job posting for whether certain words and phrases are likely to put people off applying, and thereby make the position take longer to fill, by sounding sexist or corporate jargony, flagging buzzwords like “big data” and “rockstars” in favor of “caregiver leave” and “learn new things.”
Because Internet Page 5