No single key
Cuneiform was not cracked by one person or one find. It was read again through a chain of partial breakthroughs, between about 1800 and the early 1900s, and each step depended on the one before. The script is also not “Sumerian writing” alone: the first language to be deciphered was Old Persian, which is not related to Sumerian at all. Sumerian itself, the oldest language to be written in cuneiform, was one of the last to be understood.
Copies and a clever guess
European travellers were copying cuneiform inscriptions from the ruins of Persepolis in Iran by the seventeenth century. The most useful early copies were made by the Danish surveyor Carsten Niebuhr, who visited in the 1760s and published careful drawings. Scholars could see that the inscriptions used wedge-shaped signs in several different scripts, but no one could read them.
In 1802 a German schoolteacher named Georg Friedrich Grotefend presented a proposal based on one of the three scripts, later recognised as Old Persian. He noticed that the texts were short and repeated, and he reasoned that they probably began with a formula like “X, great king, king of kings, son of Y, king”. He tested the known kings of the Achaemenid dynasty against the pattern. A text in which the father of the king had no royal title fit the line “Darius, son of Hystaspes”, since Hystaspes was not a king. Grotefend’s identification of the names Darius, Xerxes and Hystaspes gave him the values of several signs. He was a historian and a language enthusiast, not a specialist in Near Eastern languages, and some of his readings were later corrected, but his method was sound. Others, including the Danish scholar Rasmus Rask, the French scholar Eugène Burnouf and the German Christian Lassen, then extended and improved his work, partly by comparing Old Persian with Avestan and Sanskrit.
Rawlinson at Behistun
The decisive source was the Behistun inscription (also spelled Bisotun) on a cliff in western Iran. It was carved on the orders of Darius I, probably in the 520s BCE, to commemorate how he put down revolts and took the throne. Above the text is a relief of Darius with defeated rebels. The inscription is written in three languages: Old Persian, Elamite and Babylonian, a form of Akkadian.
Henry Rawlinson, a British army officer who served in Persia, copied the Old Persian text in the 1830s and the other versions in the 1840s, often working on a ledge high above the road. The Old Persian could be read with the help of Grotefend’s and others’ progress, and Rawlinson published his translation in the 1840s. Since the three versions told the same story, the Old Persian text gave a way into the others, in a way broadly comparable to the Rosetta Stone, which had allowed Champollion to read the Egyptian hieroglyphic script. The comparison is helpful but not exact. The Rosetta Stone’s Greek was already understood; at Behistun, the known-language side had itself first to be worked out.
Babylonian: a harder system
The Babylonian version was far more difficult. The script had hundreds of signs, and a sign could be a word, a syllable or a category marker, and one sign could have several readings. Early on, some scholars doubted that such a complicated system could be deciphered at all.
Rawlinson, the Irish clergyman Edward Hincks, the French-German scholar Jules Oppert and the English polymath William Henry Fox Talbot all worked on the problem. Hincks in particular understood early that the signs were syllabic and that one sign could carry several values. Names of kings, which could be matched between the Persian and Babylonian versions, provided toeholds, and the large collections of Assyrian tablets then arriving in Europe supplied vocabulary. Those collections came from excavations at Nineveh and other Assyrian sites by Paul-Émile Botta, Austen Henry Layard and Hormuzd Rassam, the work now grouped under Assyrian archaeology.
The 1857 test
By the mid-1850s there was real doubt about whether the claimed translations were genuine. In 1857 the Royal Asiatic Society in London arranged a test. A recently found Assyrian inscription of the king Tiglath-Pileser I was sent in sealed envelopes to four scholars: Rawlinson, Hincks, Fox Talbot and Oppert. They worked independently. A committee compared the four versions and found them to agree on the main points, with differences in detail. This was taken as proof that the decipherment was sound. A later public sign of the new knowledge came in 1872, when George Smith of the British Museum reported a Babylonian account of a great flood on tablets from Nineveh, which later proved to be part of the Gilgamesh tablets.
Sumerian: a language inside a language
Even as Akkadian was being read, scholars noticed that some tablets contained signs and words that did not fit a Semitic language. Word lists, set out in two columns, gave Akkadian equivalents for words in another tongue. Hincks and Rawlinson suspected that the script had been invented by a non-Semitic people who came before the Babylonians. In 1869 Oppert proposed calling that language Sumerian, from the title “king of Sumer and Akkad”.
The idea was controversial. The French Assyriologist Joseph Halévy argued that “Sumerian” was not a real language but a kind of secret writing invented by Semitic priests, and the argument continued for years. The matter was effectively settled by new material. From 1877 the French excavation at Tello (ancient Girsu) under Ernest de Sarzec produced large numbers of Sumerian texts, some of them lengthy royal inscriptions with no Akkadian at all. François Thureau-Dangin and others then worked on the grammar, and Arno Poebel’s grammar of 1923 became a standard reference. Sumerian came to be accepted as a distinct language with no known relatives.
Other cuneiform languages
Once the principle was clear, other languages followed. The Czech scholar Bedřich Hrozný announced in 1915 that Hittite, written in cuneiform on tablets from Anatolia, was an Indo-European language, an achievement now known as Hittite decipherment. Ugaritic, written in an alphabetic cuneiform script, was deciphered in 1930 within months of the first tablets being read. Not every ancient script has had such luck. The Linear A script of Bronze Age Crete, for example, remains unread.
What is still uncertain
Sumerian is understood well enough to translate most texts, but it is not read with the same confidence as Akkadian. It has no known relatives to help, the earliest texts are terse, and the spoken language died out thousands of years ago, so some vocabulary and grammar are still debated. New tablets and new editions continue to refine the picture, which is a reminder that decipherment is a process that has no final day.