The Writing Revolution

by Amalia E Gnanadesikan

  Evans came to believe firmly in the cultural superiority of the Minoans over all of their Aegean contemporaries: the Mycenaeans were warlike barbarians by contrast. And so Evans did not long entertain the possibility that the tablets he had excavated could be Greek. The more he uncovered of the Minoan age, the less possible it looked. The tablets must record the forgotten language of the Minoans.

  Looking over his tablets, Evans was able to identify three distinct forms of writing. The oldest he dubbed Cretan hieroglyphic because, like the Egyptian hieroglyphic script, it consisted of fully drawn pictograms. This was the script of the sealstones and of a small number of clay tablets. The second two scripts consisted not of full pictures but of stylized lines, and he called these Linear A and Linear B. Linear A was found on clay tablets and occasionally on objects worked in stone or metal, while Linear B only appeared on clay objects: tablets, labels and sealings, and the occasional ceramic jar.

  Cretan hieroglyphic was the earliest to develop, as the palatial centers arose at Knossos, Phaistos, and Mallia. Its birth may have been as early as 1900 BC, and it remained in use until around 1600 BC. Cretan writing did not develop in a cultural vacuum, however. The cultures of Egypt and Mesopotamia were already literate; Minoan traders would have encountered writing and brought home tales of a strange but useful technology, inspiring the development of a native script. (The Anatolians similarly invented their own hieroglyphic script and the Canaanites developed the earliest consonantal alphabet during the Bronze-Age expansion of civilization in the region.)

  Linear A appears to have derived from Cretan hieroglyphic, but was drawn with simpler lines. It seems to have been used from the eighteenth century BC to the mid-fifteenth century BC. Its birth is hard to pinpoint, however, as most of the surviving examples come from the very end of the period, when the palaces were destroyed by fire and the clay tablets on which Linear A was written were accidentally baked (none of the Cretan tablets was baked deliberately).

  Linear A was used in palaces all over Crete, but the surviving examples are few; the majority of them come from an archive of about one hundred and fifty small and poorly preserved tablets found in the royal villa of Haghia Triadha near Phaistos, baked in a fire of the fifteenth century.

  Linear B succeeded Linear A, but for many years it looked like Linear B had been used only at Knossos and on the occasional vase exported to the mainland. When the other palaces were destroyed, Knossos suffered only partial damage and was rebuilt, enjoying a renaissance between 1450 and 1350 or 1300 BC, when it was finally destroyed.

  Linear B is related to Linear A, but the precise nature of the relationship is not entirely clear. Telling them apart is not too difficult, however, as Linear B tablets tend to be neater than Linear A tablets, with the text written on ruled lines. Otherwise they are quite similar: the two scripts have about the same number of signs, many of which are the same or nearly so. But some signs of Linear A do not survive in Linear B, and some signs of Linear B have no precedent in Linear A. There are even a few Linear B signs that more closely resemble Cretan hieroglyphs than the corresponding signs in Linear A. So Linear B, though younger, may be more a sibling than a direct descendant of Linear A.

  The material in Cretan hieroglyphic and in Linear A was – and remains – relatively scanty. The number of Linear B tablets available was much bigger, however, so Linear B was the obvious place to start a decipherment. To Evans the Linear B script was clearly the product of the advanced Minoan culture in its golden age. It was a special, royal script, developed at the palace at Knossos and used during the period when Knossos was the undisputed leader of civilization not only in Crete but in the entire Aegean world. It would surely record the lost language of the Minoans.

  It was apparent, however, that neither the Linear B nor the Linear A tablets had all that much to say: judging from the length of the texts, the numerals, and the accompanying pictograms, they were clearly accounting tablets, more parallel to proto-cuneiform accounting records than to the Epic of Gilgamesh or other literary cuneiform texts.

  But what was the language of the Minoans? Eager would-be decipherers suggested connections to Basque or to Etruscan. One thing seemed obvious: it was not Greek. For one thing, the Minoans were culturally distinct from and far more advanced than the Greeks of the time; they would surely not have used the Greeks’ language. A second and more convincing reason came from a related script from Cyprus.

  The Cypriot syllabary is clearly related to Linear A and B but was used in Cyprus to write Greek from sometime after the Greek colonization of the island around the twelfth century BC until the third or second century BC, when standardization of writing spread through the Greek world in the wake of Alexander the Great (336–323 BC). It was deciphered in the 1870s using bilingual Phoenician–Cypriot inscriptions, with later confirmation from biscriptal Greek inscriptions, written both in the Greek alphabet and in the Cypriot syllabary.

  The Cypriot syllabary is a CV syllabary of the typical kind, each sign standing for a consonant–vowel sequence, except for a few signs which stand for plain vowels. Like most syllabaries (Akkadian being a notable exception), it lacks VC or CVC signs. This makes it a poor fit for Greek. Greek syllables may begin with a sequence of consonants, and may also end with a consonant, as in the Homeric name Ktesippus. As a result, Cypriot spelling was only approximate, sometimes leaving out consonants, and sometimes adding additional, unpronounced vowels. For example, words of Greek may end in the consonants r, n, or s. To write these final syllables, Cypriot used signs for re, ne, and se. Thus Ktesippus would have been spelled ke-te-si-pu-se. The sign for se in Cypriot, , looked just the same as one of the signs in Linear B. If that sign had the same pronunciation in both scripts, any Greek text written in Linear B should have many words ending in . None did. So apparently the texts were not Greek.

  The decipherment of Linear B was delayed by Evans’s slow approach to publishing the tablets. In 1909 he published the hieroglyphic texts he had found, and he planned to go on to publish the Linear A and B texts, but the more exciting work of publishing his other finds at the palace intervened, and continued to do so for much of the rest of his life. The inclusion of some tablets in the fourth volume of his Palace of Minos in 1935 brought the total number of published Linear B tablets up to only 120, a tiny amount of material considering that each tablet holds only a very brief text.

  Sir Arthur Evans died in 1941 at the age of 90, without having published the remaining Knossos tablets. They were eventually edited by a colleague and published in 1952. Even before his death, however, Evans had lost his monopoly on Linear B. In 1939 the American archaeologist Carl Blegen began digging at a site in mainland Greece that he believed to be the ancient city of Pylos, mentioned in Homer as the capital of wise King Nestor. The very first day’s digging brought to light clay tablets written in Linear B. By season’s end, it was clear that a Mycenaean king’s palace had indeed stood here, and that written records had been used in its administration. Then World War II forced a pause in the excavations. The Pylos tablets were not published until 1951, and further excavation was delayed until 1952. In 1952 a few tablets were also found at Mycenae. Since then tablets have been found at Tiryns, at Thebes, and a few at Chania, thus far the only site in Crete outside of Knossos to yield samples of Linear B.

  The mainland Linear B tablets dated from later than the Knossos tablets, to the preserving fires of a period of widespread destruction around 1200 BC, near the close of the Bronze Age. But the style of writing was the same, and its presence on mainland Greece was perplexing. Had the Mycenaeans imported their scribal class from Crete? This was not impossible, as they clearly looked to Crete for many cultural imports. Had the Mycenaean upper class learned Minoan in order to learn to write, as earlier Akkadian scribes had learned Sumerian? Only the decipherment of the tablets would yield the answer.

  The decipherment of an unknown script requires a great deal of preliminary analysis, some of which could be done even without f
ull publication of the texts. The basic subject matter of Linear B texts was pretty obvious: they were accounting tablets. A typical tablet has words written out in Linear B, followed by an ideogram, followed by a numeral (see plate 5). The use of the term ideogram is for once accurate here. In many cases these ideograms are pictographic, showing the type of entity being counted: horses (), chariots (), people (), jars (). In some cases the ideogram is stylized so as not to be recognizable by the uninitiated (e.g. , “grain”). In others it is completely abstract, being derived from the signs with which the relevant words are written (e.g. , from , a-re-pa, “ointment”). But unlike logograms in a logosyllabary, the ideograms do not appear mixed in among the syllabic signs. They only appear next to numerals, indicating what kind of thing is being counted. There is no evidence that they were even pronounced. With decipherment it became clear that they repeated information spelled out in the syllabic text of the tablets, so in all likelihood they were not pronounced. If they were, readers of the tablets would have said things like “Horses, horses 12,” repeating themselves a great deal. The ideograms served as a visual organization for the numbers recorded on the tablets, much as a dollar sign today may head up a list of numbers and identify what kind of thing is being counted.

  For twentieth-century scholars, the ideograms were of great help in guiding and evaluating decipherment attempts. Even before the decipherment, the tablets could be sorted by topic. They were clearly about food, livestock, trade commodities, weapons, and other concerns of a centralized Bronze-Age bureaucracy. Any decipherment that claimed to find in these mundane tablets an ode to the goddess Athena, say, was bound to be wrong.

  Aside from the ideograms and numerals, the core of the script consisted of about 87 signs, some uncertainty being due to the rarity of certain signs and the variation between different handwriting styles. The number of signs was a clue to the nature of the script. While the occasional language with 87 or more phonemes does exist, most languages have far fewer; so Linear B was probably not an alphabet. A logographic system with only 87 signs is even less likely. Even a complex syllabary of the Akkadian type, encompassing CV, VC, and even a few CVC syllables, would have required more than 87 symbols. In all probability, then, Linear B was a syllabary of the open-syllable CV type that Cypriot was already known to be.

  There were no bilingual texts for Linear B. Nor did history record any royal names that one could expect to find, the way Grotefend identified Darius’ and Xerxes’ names in Old Persian cuneiform. The names of mythological characters like Minos and Daedalus were unlikely to show up, and indeed have never been found on a Linear B tablet. The language, and even the language family, was completely unknown. Where, then, to start?

  One possibility was with the Cypriot syllabary, transferring the syllabic values of the Cypriot signs to similar ones in Linear B. Only seven of the signs are the same or nearly the same, though some additional ones are similar. Nevertheless, Evans himself used Cypriot to successfully read the first Linear B word, though he rejected it as a coincidence. On one particular tablet the ideograms showed horses’ heads. One was smaller and had no mane, while the other was larger and complete with mane. The smaller was accompanied by a word, , which, if supplied with Cypriot values, read po-lo. Evans was struck by the similarity to the Greek plos, meaning “foal,” but was so sure that the language could not be Greek that he did not pursue this approach. If he had, he might have noticed that the missing final s, rather than an added se, suggested that the spelling conventions of Cypriot did not apply to Linear B, undermining the argument that it could not be Greek.

  The similarities with Cypriot were not strong enough to allow a decipherment, and those who worked on the decipherment wisely did not assume that signs that looked similar in Cypriot must have the same reading in Linear B. Comparisons between the Greek and Roman alphabets, where P stands for [r] in Greek but [p] in Latin, for example, show the dangers of such an approach.

  The American Alice Kober’s particularly clear-thinking approach was to look for evidence as to what kind of language the tablets recorded. Working between 1943 and her death in 1950, she was to make the most progress before the actual breakthrough. Scrutinizing the published texts, analyzing the syllabically written words in the context of the ideograms and numerals, she looked for evidence of how the language assembled its words. Luckily (and unlike the early alphabetic inscriptions of later Greek), it was clear where one word ended and another began, the break between one word and the next usually being shown by a short vertical line.

  Some tablets recorded several numerals, with a total at the bottom of the tablet. Kober found that the word preceding the totals had two forms, one used for men and a certain class of animals, and another, with the same first syllable but a different final syllable, used for women and another class of animals. The language of Linear B therefore had gender, dividing the world up into masculine and feminine things, like Spanish, French, or ancient Greek. In fact, the particular way the masculine and feminine forms were related (by a difference in the final syllable, rather than by the presence or absence of a suffix) made it very likely that Linear B recorded a language of the IndoEuropean family, to which Greek, Latin, and even English belong.

  Perhaps most significantly, Kober found sets of words whose endings varied in regular ways. These came to be called “Kober’s triplets” (see figure 6.1). Kober rightly concluded that these words shared a core morpheme but had different endings depending on their grammatical form. In English, for example, Canada is the name of a country. But the AD jectival version of the word is Canadian, whose meaning is obvious from the combination of Canad- and -ian, but is not formed simply by adding a suffix -n to the end of Canada. Bermuda and Bermudian, and Argentina and Argentinian, form pairs according to the same pattern. The language of Linear B words formed triplets rather than pairs.

  Figure 6.1 Some of Kober’s triplets, which she took as evidence for inflection in Linear B. The triplets were foundational to Ventris’s logical grid of consonants and vowels and also provided the key to the phonological decipherment of Linear B when Ventris guessed that they recorded the names of places in Crete. The ones at the top refer (from left to right) to Amnisos, Knossos, and Tylissos. At bottom left is Phaistos, at right Lyktos.

  Kober’s triplets revealed something even more important. If Canada and Canadian are divided syllabically, you get ca-na-da and ca-na-di-an. There are two different third syllables in these words, da and di, but they begin with the same consonant. By analogy, it was possible to predict that certain syllables shared a consonant. In the triplet , , and , and were likely to have the same first consonant, as were and . By comparing one triplet set to another, it was possible to predict that certain other syllables shared a vowel. Bermu-di-an and Ar-gen-ti-ni-an have the same vowel in the second-to-last syllable, as they both have the -ian suffix. Similarly, in the suffixed forms and , and both came before and were therefore likely to share a vowel. For every pair of signs for which Kober hypothesized a shared vowel or consonant, she was correct. She also pointed out, wisely, that Linear A probably recorded a different language than Linear B. For all the similarity of the two scripts, the Linear A tablets record not a single word that is also found in Linear B.

  At this point leadership in the decipherment passed to a young British architect by the name of Michael Ventris. Ventris had been fascinated by the mystery of Linear B since 1936, when at the age of 14 he had met Sir Arthur Evans. His own pet theory was that the language was Etruscan, and he held to this idea until the final steps of his solution showed him that it could not be. Ventris followed up on Kober’s work, organizing the signs of Linear B into a grid, with 15 rows of signs that he believed shared a consonant and 5 columns of signs that he believed shared a vowel. Yet it was his ability to entertain hunches, as well as his painstaking work producing ever more complete versions of his syllabic grid, that led him to a solution.

  The Pylos tablets were published in 1951, finally making available
enough data to allow a true decipherment. Ventris quit his job to focus on Linear B. He filled in his grid with instances where signs probably shared a consonant or a vowel, gathering evidence from variant spellings of what were probably the same word; from still-visible erased signs (indicating a corrected spelling mistake, and thus perhaps a similarity between the incorrect and the correct spelling); from inflections, including the kind in Kober’s triplets (of which many more examples were now available); from masculine and feminine endings; and from singular and plural forms of words (identifiable by the accompanying numerals). He noted that three signs favored the first position in a word. These were probably vowels – plain V rather than CV signs – which are especially needed at the beginnings of words. Ventris was able to build his grid with a high (but not perfect) degree of accuracy, based on the relative values of the signs, with as yet no good idea of what the actual phonological values of any of the signs were. In other words, he could have told you that and shared a consonant, but he could not have told you what that consonant was.

  With this careful groundwork laid, what was still needed was an inspired guess. Ventris had noticed that the triplets Kober had found in the Knossos tablets did not occur in the Pylos tablets. Might they then be the names of places on Crete, with suffixes on adjectival forms (like Canada, Canadian)? What places were they likely to be? Some Cretan place names known from classical times were clearly of ancient, pre-Greek origin, as they share a suffix that is meaningless in Greek: Amnisos (the harbor nearest Knossos), Knossos, Tylissos, Phaistos, Lyktos. One of Kober’s triplets began with one of the signs Ventris had identified as a vowel. The uninflected form had four syllables: . Might this be Amnisos, spelled syllabically as a-mi-ni-so? At this point Ventris called upon some similarities with the Cypriot syllabary. The Linear B sign was very similar to the Cypriot sign for na, . If the similarity could be trusted, and if really was a, then should be in the column headed by . It was. This identified the -a column, and by implication the n- row (for a grid layout of the signs of Linear B, see figure 6.2). Another sign similar in Linear B and Cypriot was ti, . If the signs really were the same, this identified the t- row and the -i column. The third syllable of the proposed a-mi-ni-so, , should be in the -i column and the n- row. It was. The second syllable, , should also be in the -i column, and indeed it was. This suggested an identification of the m- row. Guessing that the fourth syllable, , was so allowed more of the grid to be filled in.