Cuneiform, and the men who guessed

For somewhere near two thousand years, nobody on earth could read the world's oldest writing. The wedge-marks were everywhere in the ruins of Persia and Mesopotamia, pressed into brick and cut into cliff faces, and they kept their silence through the entire life of classical scholarship. The recovery, when it came, was not one discovery but a relay of guesses, each solving a script that unlocked an older and harder one; and it ended with a public examination, four scholars, and sealed envelopes.

The relay starts with a survivor. Carsten Niebuhr was the last man standing of a Danish expedition to Arabia that killed every other member; he came home alone, and in March 1765 he stopped at the ruins of Persepolis and copied its inscriptions with a draughtsman's patience. His copies, published in 1778, were accurate enough to work from, which no earlier traveller's had been, and he noticed the essential thing: the walls carried not one script but three, side by side, in increasing degrees of complication. The simplest used few enough signs that he took it for an alphabet. Nobody could read a word of any of them.

Small footholds followed. Oluf Gerhard Tychsen spotted that a recurring oblique wedge in the simplest script was a word divider, then dated the whole corpus to the wrong dynasty. Friedrich Münter corrected the dating, placing the ruins with Cyrus and his successors, and guessed that one frequently repeated word ought to mean "king". And Antoine Silvestre de Sacy, working on the much later inscriptions of the Sasanian kings, recovered their formula: X, great king, king of kings, son of Y. That formula turned out to be the key that fits the lock two dynasties earlier.

The man who turned it was not an orientalist. Georg Friedrich Grotefend was a twenty-seven-year-old classics teacher in Göttingen who, the story goes, took the problem on as a wager; the tale rests on later recollection, so hold it lightly. What is certain is his method, because it needed no Persian at all. Assume the Persepolis texts are royal. Assume they follow de Sacy's formula. Then in two of Niebuhr's inscriptions, the name standing where the king's father should be in one is the name standing where the king should be in the other; a father and son who both reigned. And in the first inscription the grandfather's name carries no royal title; a king whose father was no king. The Achaemenid king-list, known from Herodotus, offers exactly one fit: Darius, son of Hystaspes, succeeded by Xerxes. Grotefend pinned the names, read sound values off them, and presented the result to the Göttingen academy in September 1802. He had correctly identified perhaps a third of the signs he worked with; wrong on much of the rest, right where it mattered. The academy declined to publish. The full text of his papers finally appeared in 1893, forty years after his death.

The simplest script, Old Persian, gave way over the following decades; it turned out to be a semi-alphabetic system of thirty-six phonetic signs, invented under Darius and abandoned within two centuries, the youngest and easiest member of the whole cuneiform family. The real prize, and the real problem, was the many-hundred-sign script of Babylon. Reading it required a long text in both scripts, and the longest in existence was six hundred kilometres away, carved high up a cliff at Behistun in western Iran, where Darius had posted the official account of his seizure of the throne in three languages, positioned to be seen and not reached.

Henry Rawlinson, an East India Company officer posted to Persia, spent parts of a decade getting it down: climbing to the narrow ledge below the Old Persian panels through 1836 and 1837, returning in 1844 to recopy and take paper squeezes, carving his own name below the inscription while he was at it. The Babylonian panel defeated even him; he studied it by telescope, and in 1847 the squeezes were finally taken by a local climber, in Rawlinson's telling a "wild Kurdish boy" who crossed the rock face and rigged a seat from a rope; the story is Rawlinson's own, and its details vary with his retellings. His Old Persian text and translation began appearing in 1846. The trilingual was in hand.

What the Babylonian script then demanded was a different order of guessing, and much of it was done far from any ruin, in an Irish rectory. Edward Hincks, a Church of Ireland clergyman in Killyleagh who never in his life saw Mesopotamia, worked out the character of the thing sign by sign: that the script was syllabic rather than alphabetic; that a single sign could carry several different readings depending on context; that certain signs were never pronounced at all, but stood as silent classifiers marking the word beside them as a god, a city, a man. Each of these looks, in hindsight, like the discovery of a fact about writing itself. Hincks also saw early that the language behind the script was Semitic, kin to Hebrew. Rawlinson published in parallel, claimed his independence loudly, and history took him at his word for a century; the modern reassessment, resting partly on what the British Museum let him see of Hincks's notebooks, is considerably kinder to the rector.

There was one more door at the bottom of the stairs. As the Babylonian texts opened, it became clear that the script had not been designed for a Semitic language at all; the fit was too poor, the sign-values too layered. Hincks said it first, in 1850: the inventors of cuneiform spoke something else. The something else emerged from bilingual school tablets, ancient dictionaries compiled by Babylonian scribes for a language already dead in their own day. Oppert gave it its name in 1869: Sumerian. It is related to nothing else we know. For decades one eminent scholar, Joseph Halévy, insisted it was no language at all but a priestly cipher, and was wrong at length. The last language recovered was the first ever written; the relay had run the whole way back.

It is worth saying what made the run possible. Not one of the guesses was unconstrained. Grotefend's names had to survive a king-list; Hincks's syllables had to survive every new tablet; the 1857 test put the whole method in front of a jury on purpose. Guessing, checked ruthlessly against material that cannot be argued with, is simply philology working at full stretch; and it recovered, from wedges in clay, the accounts, letters, laws and complaints of the first people who ever wrote anything down.