The Riddle of the Labyrinth by Margalit Fox

wrote to Mary Swindler, the editor of the American Journal of Archaeology , in January 1941.
    Where Evans’s approach to Linear B was scattershot, impressionistic, and anecdotal, from the start Kober imposed more rational methods. Her first order of business, starting in the 1930s, was frequency analysis: the creation of statistics “of the kind so successfully used in the deciphering and decoding of secret messages,” as she wrote, for every character of the script.
    Anyone who has solved a Sunday newspaper cryptogram has met frequency analysis head-on. At its simplest, it entails pure counting, with the decipherer tabulating the number of times a particular character appears in a particular text. If the text is long enough, the frequency count for each letter should mirror its statistical frequency in the language as a whole. It was frequency analysis that let Sherlock Holmes decipher this secret message, one of several at the center of Sir Arthur Conan Doyle’s story “The Adventure of the Dancing Men.” (The font used here differs somewhat from the original, but the message is the same):

    As Dr. Watson and a provincial police inspector look on, Holmes elucidates his method. “As you are aware,” he says, “E is the most common letter in the English alphabet, and it predominates to so marked an extent that even in a short sentence one would expect to find it most often.” To Holmes, it was immediately apparent that the character, the most frequent in the cipher, stood for “e.” As for the rest of the alphabet, he continued, “Speaking roughly, T, A, O, I, N, S, H, R, D, and L are the numerical order in which letters occur.” (The message, quickly unraveled, read: “Elsie. Prepare to meet thy God.” Happily, Elsie did not.)
    Every language has its own characteristic letter frequencies, and for decipherment, this fact can be telling. In German, the eleven most frequent letters, in descending order, are e n i r s t a h d u l . In French, they are e a s t i r n u l o d . Thus, for a simple substitution cipher like the one above, a moderately skilled investigator can use frequency analysis first to identify the underlying language (assuming it is one whose letter frequencies are known) and then to crack the cipher itself.
    For frequency analysis to work properly, the text of the cipher must be long enough to provide a statistically significant sample. And that, for Kober and other investigators of Linear B, was precisely the problem: Evans, resolute as ever in old age, had continued to sit on his data. In the early 1930s, when Kober first turned her attention to the Cretan scripts, the only inscriptions to which anyone had access were the tiny handful Evans had published in Scripta Minoa in 1909, plus the small set published covertly by the Finnish scholar Johannes Sundwall—the “thesaurus absconditus,” scholars called it. The two sets together totaled fewer than one hundred inscriptions, less than one-twentieth of what Evans had unearthed at Knossos.
    With so little text available, how can a decipherer even begin? Kober began by looking for ghosts.
    EVERY LANGUAGE GLIMMERS with sparks of earlier ones. These sparks—a word, a place-name—are the residual traces of languages spoken before, often long before, in the same part of the world. Though tiny, the sparks can illuminate a history of invasion, conquest, trade, and the wholesale movement of populations. In the West Germanic language known as English, we can discern Julius Caesar’s invasion of Britain in the first century B.C. from linguistic survivals like wine (from Latin vinum ) and anchor (Latin ancora ) that remain in use today. We see the enduring presence of the Celts, who inhabited Britain before the Romans and for some time afterward, in place-names like Cornwall, Devon, and London. We also see the legacy of the Viking conquests of Britain toward the end of the first