THE CODEBREAKERS

by DAVID KAHN

  The error of her ways was demonstrated in a 1940 article in the American Journal of Archaeology by one M. G. F. Ventris, who, whoever he was, was evidently not a professional archeologist, since, where most contributors listed their university affiliations, he gave only “London.” Yet his article was good enough to be published. Ventris cleared the field of opposition views—“The theory that Minoan could be Greek is based of course on a deliberate disregard for historical plausibility, and the wonder is that the Greek readings have been got into publishable form at all”—before presenting his own case for a similarity with Etruscan. He based it upon the lack in both Etruscan and the Cypriote syllabary of the voiced stops /b/, /g/, /d/ and the apparent derivation of Cypriote from Minoan. He deciphered some Linear B names with the help of the syllabary and obtained some “name-radicals,” which did not add much except confusion. Concluded Ventris bravely: “It [the decipherment] can be done.”

  Meanwhile, Bedřich HroznÝ, who had disposed of the Indus Valley hieroglyphics to his satisfaction, knifed through Linear B with a facility that made everyone who had been stymied by its difficulties look like fools. He assigned the phonetic value /ha/ to a Linear B sign, for example, on the basis of resemblances that he saw between it and a Hittite sign for hà, an Egyptian sign for he, a Sabaean for h, a Carian and Etruscan for kh, and a Phrygian for x. He related another sign to a value deduced from his Indus Valley decipherment. The solution consisted of a hodgepodge of words from a mixture of different actual and derived languages. Critics condemned it on methodological, philological, and evidentiary grounds. “When the decipherer is as thoroughly acquainted with as many languages as HroznÝ certainly is,” one critic observed gently, “the range of possible satisfactory combinations of sound and sense is very large, and few inscriptions will seem entirely without sense.” Most importantly, HroznÝ’s readings did not correspond to what the tablets appeared to be talking about.

  At about the same time, Vladimir Georgiev of the University of Sofia offered an 81-page decipherment. He wrote that his study of Aegean place-names and Greek vocabulary had established that the Linear B language was an unknown Indo-European tongue which he named Pelasgian or Eteocretan. His readings seemed arbitrary, however, and did not secure conviction.

  Essays more modest appeared from time to time. In 1950, the German scholar Ernst Sittig, who had served in the German Foreign Office crypto-logic section from 1919 to 1924, assumed that the language underlying some non-Greek inscriptions written in the Cypriote syllabary was related to Minoan. He then matched these signs to Linear B on the basis of resemblances in form and in frequency, and announced the identification of 14 signs. With these as a start, he began work to recover the Minoan language.

  The most valuable of the limited studies was a series of articles by Dr. Alice B. Kober, assistant professor of classical studies at Brooklyn College. In 1944 she presented a close textual analysis of tablets with an adze ideogram, and in 1945 pointed out that the final signs in words on ten “chariot” tablets varied. As Evans had suggested ten years earlier, she concluded that “it is highly probable that the language of the Linear Class B documents was inflected.”

  In an inflected language, changes in the form of words—usually endings—indicate differences in tense, gender, number, person, case, and so on. Only a few such changes, or inflections, survive in English. An -s on nouns marks a plural. An -ed on verbs indicates past tense. Inflections are much more common in older languages, as anyone who has grappled with Latin grammar knows. Thus, where English would use the simple form earth in all cases, Latin declines it from terra to terrae, terrae, terram, and so on, depending on whether it is a singular noun in the nominative case, the genitive case, the dative case, or some other case. The part of the word that does not change—terr—is the stem.

  “If a language has inflection,” Miss Kober wrote, “certain signs are bound to appear over and over again in certain positions of the written words.” She detected such repetition, though she conceded that “the types of inflection used, and their significance are still unknown.”

  The Linear B nouns that Alice Kober used in her original analyses

  The next year she identified the signs that constituted the inflections. She began by assuming that the words in a tablet headed with the ideogram for “woman” were all nouns (probably names) and all in the same case. Then she postulated that “if a certain sign or group of signs occurs regularly or with fair frequency as a word ending in a given inscription … this ending is usual for … the particular case.” In Latin, the -ae of the genitive singular will recur in terrae, fossae, barbae, and so on. She found a sign that recurred thus frequently as a word ending. It looked like a ladder and was referred to for typographical convenience as “7.” She labeled it the ending for Case I. She could not tell, of course, whether the case was nominative, accusative, or what. Next she looked in other tables for the same words with a different common ending. She could recognize the “same” words by the invariant stem. She found another ending, which looked somewhat like a 5, referred to as “40,” and labeled it the Case II ending.

  When she had done this for all the common endings, she found several nouns that were declined in three cases. For her analysis, she in effect picked out a pair that exhibited some puzzling characteristics and concentrated on them, hoping that explaining the characteristics would help solve Linear B. The two nouns thus chosen may be tagged by letting JK represent the signs of the stem of one and LM the signs of the stem of the other. She set them out in two paradigms, each listing all forms of one word:

  Case 1 J K 2 7 L M 36 7

  Case II J K 2 40 L M 36 40

  Case III J K 59 L M 20

  Miss Kober then dared a conjecture that might explain these variations. Suppose, she said, that the signs of the Linear B syllabary could represent only either pure vowels or syllables of consonant-vowel formation. She based this assumption upon the resemblance of the Linear B to the Cypriote syllabary, which could express sounds only in that very way—syllables like a and da permitted, syllables like ad and dap and lone consonants like d excluded.

  Suppose further, she said, that both the JK and LM stems ended in consonants. This assumption was justified; most stems in most languages seem to end in consonants—hom for Latin homo, for example. The end of the stem would be followed by the beginning of the inflection, and in a Cypriote-like syllabary, the consonant would have to be followed by a vowel. Thus the syllabary would link together in a single sign the consonant of the stem ending and the vowel of the inflection’s beginning—the m and the o of homo into mo, if Latin had been written in that syllabary. Such a sign would straddle or bridge the natural division between stem and inflection. It would stand with one foot in the stem and the other in the inflection, the first foot being a consonant, the second a vowel. It may be called a “bridge sign.”

  Now in a paradigm, the first vowel of the case ending often varies as part of the variation that differentiates one case from another. Thus, in the Latin paradigm dominus, domini, domino, the first vowel of the inflection is successively u, i, and o. Hence, in a Cypriote-like syllabary, the bridge signs that would incorporate these varying vowels would themselves vary, because nu, nu, no would necessarily have different signs. Miss Kober observed this phenomenon in the variation between signs 2 and 59 in the JK noun and between signs 36 and 20 in the LM noun. She therefore regarded them as bridge signs. But—concentrating on JK—signs 2 and 59 each stand with one foot on the unchanging final consonant of the stem. Thus both these signs begin with the same consonant. For generality, Miss Kober illustrated this principle with an Akkadian noun, sadanu, whose stem is sad- and whose case endings are -anu, -ani, and -u:

  Case I J K 2 7

  sa da nu

  Case II J K 2 40

  sa da ni

  Case III J K 59

  sa du

  Miss Kober did not suggest that these were the actual meanings of the Linear B signs. She simply wanted to demonstrate
how signs 2 and 59 shared the fixed consonant of the stem. By the same reasoning, signs 36 and 20 of the lm noun began with a consonant in common. In neither noun did she know what the consonant might be. (She could not draw any conclusions about signs 7 and 40, for though in Akkadian they happened to have the same consonant, in Linear B they might not. Case II might be something like sadalo.)

  The Brooklyn College scholar thus ascertained that some signs shared a common consonant. She thereby drove the thin edge of a wedge into the theretofore unbreached façade of Linear B. In her next move, Miss Kober widened this crack into a substantial fissure.

  She cross-compared the JK and LM nouns. She recalled her original search and conclusions: JK and LM had the same signs at their tails and so had the same case endings. She focused on Case I. Since both words had the same case endings, both contributed the same vowel to their respective bridge signs—2 in JK, 36 in LM. But if the vowels were the same, why were the bridge signs different? Because, she answered herself, JK was a different word from LM, the different words had different stems, and the different stems furnished different final consonants to the bridge signs. Therefore they differed. But the vowel did not change. It remained the same in both bridge signs. And so Miss Kober ascertained two signs that had a vowel in common.

  The situation can be depicted with the made-up word petanu, of the same declension as sadanu, and consequently with the same endings -anu, -ani, and -u.

  Case I J K 2 7 L M 36 7

  sa da nu pe ta nu

  Case II J K 2 40 L M 36 40

  Case III J K 59 L M 20

  sa du pe tu

  Even though signs 2 and 36 differ because they have different stem consonants, they have the vowel of the case ending in common. Likewise 59 and 20 have a vowel in common.

  Miss Kober had thus discovered some Linear B signs that had vowels in common and some that had consonants in common. Some signs belonged to both groups, and when this occurred she could arrange them in a two-dimensional pattern, with the signs sharing the same vowel in a single row and those sharing the same consonant in a single column:

  Her method was ingenious, rigorous, and powerful in the extreme. It precluded wild guesses as to the meaning of a sign, for any phonetic assumption would have to validate itself with the consonants of its column in the pattern and the vowels of its row. In other words, if du looked good for 59, d would have to make sense as a consonant wherever 2 appeared, and u as a vowel wherever 20 appeared. At the same time, it would suggest new values. If du was 59, then the insertion of d? wherever 2 appeared, as in ??-d?-nu, might suggest that 2 was da, giving a new vowel value. Then 36 would have to be ?a, and this in turn might suggest ta to make ??-ta-nu.

  Miss Kober purposely refrained from the critical step of assigning phonetic values to the signs because she felt it unwarranted with the paucity of material then available. But she wrung a few more details from the tablets, such as the demonstration that the two forms of a two-sign word at the foot of several lists represented masculine and feminine forms of “total.” Most important, by 1948 she had extended her consonant and vowel equivalents from four signs to ten, which she arranged in a “tentative phonetic pattern” two vowels deep and five consonants wide. Two years later, aged 43, she was dead of cancer.

  A few months before her death in May, 1950, she had received a questionnaire on the Linear B problem from Michael Ventris, who had propounded the Etruscan theory in the American Journal of Archaeology in 1940. The ten-year hiatus in his work was the result of interruptions by World War II, in which he served as a navigator in a Royal Air Force bomber, and by his studies at the Architectural Association School in London, from which he was graduated with honors in 1948. For Ventris was an architect, not a professional scholar, and he was not yet 30. He had written his 1940 paper when he was only 18, a fact that he had carefully concealed from the editor and that makes the article’s acceptance all the more impressive.

  He was born on July 12, 1922. His father was a British Army officer in India, his mother a beautiful woman who brought Michael up in an artistic atmosphere. He himself was uncommonly handsome. He went to school in Switzerland and then won a scholarship to Stowe School in England. His aptitude for languages was marked: he taught himself some Polish (his mother was half Polish) when he was 6, and as a young man learned enough Swedish in a few weeks to get a temporary job in Sweden. He had been taught in French and German in Switzerland, and had studied Greek at Stowe. He combined a remarkable visual memory with a good ear. As an architect, he worked for a while designing schools for the Ministry of Education, and in 1956 won the first research fellowship awarded by the Architects’ Journal. His wife, also an architect, designed a modern home for them and their two children. By all accounts he was charming and modest, serious yet with an occasional flash of gaiety, affable, able to explain things simply, and brilliant.

  Ventris’ interest in the Linear B problem had been roused when, at 14, he heard Sir Arthur Evans himself lecture on fabulous Crete and its mysterious writing. At that impressionable age, when so many lifetime enthusiasms are formed, he took up the challenge of the undeciphered script, reading the literature and, later, corresponding with the experts. The publication of seven newly discovered tablets in 1950 encouraged him to resume his analyses, beginning by determining the “state of the art.” The questionnaire that he sent Miss Kober in 1950 also went to eleven other scholars who he knew were working on Linear B. Ten of the twelve answered. HroznÝ, then past 70, did not, nor did Miss Kober, who believed—with some justification—that discussion of unproved theories is a waste of time. Ventris circulated the replies, which summarized what was known about Linear B 50 years after Evans’ discovery of the first tablets and which has come to be called the “Mid-Century Report.” The consensus was that the underlying language was probably related to Hittite; a minority, including Ventris, held that it was more closely related to Etruscan.

  In 1951, 556 of the Linear B tablets that Carl Blegen had found at Pylos in 1939 were published, thus at one stroke quadrupling the quantity of text available for study—the Evans tablets still not having been released. The publication of Blegen’s find was supervised by one of Blegen’s students, Emmett L. Bennett, Jr. Bennett, who had worked as a cryptanalyst during World War II, had written his doctoral thesis on “The Minoan Linear Script from Pylos.” Like Miss Kober, he proceeded with caution; progress was slower than if he had attacked the problem wholesale with sign substitutions and the like, but it was substantially surer. In a 1950 article, he clarified the numerical and mensural systems of both Linear A and B. But his greatest contribution was the establishment of the signary by recognizing variant forms. This first step, which can be quite difficult in an unknown script—and sometimes is not easy with just an unfamiliar handwriting—supports all the rest; it is therefore essential, but no one before Bennett had really done it. Bennett also classified the signs according to their form and established an order which others numbered to make it easier to cite the signs in print.

  By then Ventris was circulating Work Notes averaging eight pages each that he duplicated and mailed at his own expense to two dozen interested scholars. These notes reported his theories, comparisons, wild surmises. He was, in a sense, working in public, allowing each of his steps to be seen and criticized by his colleagues, and—what is frequently important among scholars, who are rewarded in fame and honor and not in cash—risking that his suggestions might touch off a train of thought in a colleague’s mind, letting him achieve the final solution. The first Work Note, mailed out in January, 1951, reviewed the evidence for inflection and for Miss Kober’s phonetic pattern. Ventris adopted it—though he placed the vowels at the top, the consonants at the side—and called it a “grid.” He drove nails into a board and hung tags on them marked with the Linear B signs.

  The second Work Note suggested that a button-like sign represented an enclitic “and”—a conjunction like the Latin suffix “-que,” which was tacked onto the end of words,
as in “Senatus Populusque Romanus,” the full form of SPQR, “the Senate and the people of Rome.” Several succeeding Work Notes tested possible parallels with a postulated Aegean language or with Etruscan, which Ventris still regarded as the probable answer. And all this time he was slowly filling in the grid by repeating Miss Kober’s technique of comparing words to determine signs sharing the same vowels and consonants. He moved the tags from one nail to another as he tested assumptions, noting whether a sign hung in a certain column seemed to have the same vowel as the signs already there.

  Work Note 8 tabulated the frequencies of each sign as initial, final, or medial. The enormous frequency of three signs at the beginning of words—one looking like a double ax, the second a throne, the third like an A with an extra bar—suggested that they might be pure vowels. In languages written syllabically, statistics showed, the pure vowels have the highest initial frequencies. Ventris thought, as others had privately, that the double ax represented a and the throne i. The assignments of pure vowels were independent of the grid; they did not affect it nor it them.

  The next Work Note set forth evidence that certain signs represented similar sounds. Ventris observed that certain words exhibited slight differences in spelling and, because these words occurred in identical sentences, he concluded that the differences represented, not inflection, but slight variations in pronunciation. One scribe might write “father,” another—from the Knos-sos equivalent of Brooklyn—might set down “fadder.” No dictionaries existed to standardize spelling; the scribes wrote what they heard. Ventris, coming across such variations in identical contexts, assumed that /th/ and /dd/ represented similar sounds. He could then place them in either the same column or the same row of the grid, depending on whether the consonant or the vowel varied, information that came from other sources. These spelling variations greatly expanded the grid.