Your Words Are Your Fingerprint

Two statisticians once identified a stranger by counting how often he used the word "upon."

The stranger was James Madison. The year was 1964. Frederick Mosteller and David Wallace had been handed the twelve Federalist Papers whose authorship — Madison's or Hamilton's — historians had argued about for a century and a half. Mosteller and Wallace ignored the arguments. They ignored the topics, the politics, the sentiments, the vocabulary a schoolteacher would notice. They counted, instead, the little words. "Upon." "While." "By." "To." The tiny hinges of English prose that nobody chooses on purpose. Madison, it turned out, used them at rates so consistent, and so different from Hamilton's, that the odds against a wrong attribution ran into the dozens on the weakest of the twelve papers, and into the thousands or more on the rest. Publius was Madison. The little words gave him up.

Attribution is the smaller point. The part of you a stranger actually meets when they read a letter from you is not your name, your face, your job, or the topics you happen to raise. It is the way you use the word "the." It is whether you write "I just wanted to say" or "I wanted to say." It is the length of your sentences, the beats between your commas, whether you reach for "we" or "you" or "one" when you are trying to be honest. These are choices you do not know you are making. English has fewer than five hundred function words — pronouns, articles, prepositions, auxiliary verbs, conjunctions — and, as the psychologist James Pennebaker points out in The Secret Life of Pronouns (2011), those five hundred account for more than half of everything any of us speaks or hears or reads. They are the water. Each of us stirs it in a way that is, to a startling degree, ours.

Which would be a party trick — Rowling unmasked as Galbraith by a Duquesne professor's software in 2013, a courtroom expert witness produced by counting prepositions — except that the fingerprint turns out to be social. In 2011, a team led by Molly Ireland and James Pennebaker took transcripts from forty four-minute speed dates and measured, for each pair, how closely the two speakers' function-word rates had drifted toward each other over four minutes. They called it language style matching. Pairs above the median were roughly three and a half times more likely to want a second date. In a second study, the same measure applied to eighty-six couples' instant-message logs predicted, three months on, who was still together. The predictor was not what the couples said. It was whether their small, unconscious words had leaned toward one another.

What gets counted

Two people describe the same rainy commute. One writes: The train was late again. I stood on the platform. Everyone looked tired. The other writes: So we were all just standing there, and the train just wasn't coming, and everybody kind of looked at each other. Both accurate. Both about the same event. Neither writer chose any of it.

The technical name for this is stylometry. Its unit of analysis is not "voice" or "tone" — words too vague to measure — but frequencies. Pennebaker's Linguistic Inquiry and Word Count dictionary splits the language in half. On one side, content words: nouns, verbs, adjectives, the things a schoolteacher would underline. On the other, function words: pronouns, articles, prepositions, auxiliary verbs, conjunctions, negations. The content words tell you what a person is talking about. The function words — six or seven categories the eye glides right over — tell you who is talking.

People who were only ever their sentences

Nothing modern about this. Rilke exchanged ten letters with the nineteen-year-old cadet Franz Xaver Kappus between 1902 and 1908, never met him once, and produced what became, three years after Rilke's death, Letters to a Young Poet. "Be patient toward all that is unsolved in your heart," he wrote to Kappus in July 1903, to a stranger whose face he would never see.

Or something less rarefied. In October 1949, Helene Hanff, a broke freelance writer in New York, wrote to Marks & Co, an antiquarian bookshop at 84 Charing Cross Road in London, asking after a used Hazlitt. Frank Doel, the chief buyer, wrote back with grave English politeness. They corresponded for twenty years. She sent tinned ham and nylons to rationed London; he sent her the books she could not find in Manhattan. When Doel died of peritonitis in December 1968, his widow wrote to tell Hanff. Hanff still had not seen the shop. She published the letters in 1970 as 84, Charing Cross Road. Twenty years, no meeting, a real friendship that turned into a book.

The obvious objection

The obvious objection: this is projection. In a letter you are not meeting the writer, you are meeting a screen onto which you throw whatever you want. Joseph Weizenbaum built ELIZA in 1966, a program that did nothing but mirror your sentences back at you as questions, and users at MIT began confiding in it — his own secretary, who had watched him build the thing, asked him to leave the room so she could talk to it privately. If people will bond with a script, surely they will bond with anything blank enough to reflect them.

The objection has a real target, and misses it. Joseph Walther's hyperpersonal model, published in 1996 in Communication Research and still the standard reference in the field, does not deny that people idealize a stranger they cannot see. It says the idealization is one of four mechanisms — the writer edits, the reader idealizes, the channel allows composition, the two lean into each other's compositions in a feedback loop — that make written, asynchronous exchange produce, under the right conditions, more felt intimacy than face-to-face conversation of the same length. Nothing in the model requires the idealization be false. It requires only that the writing be honest.

And projection alone cannot explain the Ireland data. If a reader were only painting the writer, there would be no reason two people's small-word rates would move toward each other in a way that predicts second dates and, three months on, whether the couple is still together. The rates move. The prediction holds.

The honest limit

None of this is a substitute for a face across a table. And John Suler, writing in 2004, was clear about what the same anonymous, asynchronous channels also release: benign disinhibition — unusual kindness, unusual candor, generosity between strangers — and toxic disinhibition — harassment, cruelty, contempt. Both grow from the same soil. Which one you get depends less on the medium than on what the medium is built to reward. A comment section built for outrage will get outrage. A mailbox built for a letter will, more often than not, get a letter.

What the mailbox is for

Two four-hundred-word letters take longer to write than two hundred texts. You will not be told, when you send one, whether the other person has read it. A profile photograph tells a stranger what you looked like on a Tuesday when you liked the light. A letter tells them something else. You will not know the other person's name for a long time. Sometimes never.

There is a place, still, for a room built that way. A page, a stamp, another page. No profile to keep current. No read-receipt hovering under what you sent. Just a letter, on purpose, to a stranger — and, at the other end, a person whose thes and justs and small hesitations have, over long enough, their own shape.

So: the person one writes like and the person one looks like are not, it turns out, the same person. A letter does not remove the writer. It removes some of what usually stands in front of the writer. Call that anonymity if you want. It is closer to the opposite — the self, minus the costume.