Who Wrote It?

Attributing an anonymous text to an author is an old problem that has recently acquired new tools and, with them, a new kind of overconfidence. The traditional method was internal evidence: a distinctive image, a private allusion, a habit of argument. The modern method counts words - not the striking ones, which a writer chooses deliberately, but the invisible ones, the rates at which an author reaches for "of" rather than a possessive, or begins a sentence with "but".

[A] The reasoning behind this is sound. Function words are used below the level of conscious attention, they are frequent enough to be counted reliably in short texts, and they are hard to imitate deliberately because a forger does not know which of them matter. [B] Where the method becomes unreliable is in what it is asked to do. Deciding which of two known candidates wrote a text is a comparison; deciding whether an unknown person wrote it is a claim about every writer who has ever lived, and no set of counts can support that. [C] The difference is often lost when a result reaches a newspaper. [D] A responsible study reports that the text resembles one candidate more than the other, which is a genuine finding and a much smaller one than the headline it produces.