4 The theory that Minoan Linear B is an early dialect of the Etruscan-Lemnian family is to be constantly tested against apparent meanings and inflections – but not to the exclusion of possible analogies with other language groups.
5 Each of the 79 ‘syllabic’ signs…is purely phonetic in every sign group in Bennett’s index, even if some sign groups are occasionally abbreviated.…This means that the initial analysis can take no account of the shapes of signs used in sign groups. [Ventris mentions 79, not 89, signs, because he has excluded ten signs as being too rare for useful analysis.]
6 Identifications of Minoan signs with Cypriot syllables must be confirmed by linguistic evidence, and not based upon superficial resemblances, however methodically these are applied.…
7 The simplest, most mundane and least surprising explanation of any inscription is likely to be the correct one. Both the archive tablets and the Minoan inscriptions written on other objects are likely to have the same general content as similarly written inscriptions from cultures near in time and space…[e.g. Mesopotamia].
8 We may never succeed in fully interpreting these tablets, since we should probably not be able to understand more than a small part of the Minoan vocabulary even if we knew how it was pronounced. Our limited objective must be to give phonetic values to the majority of the signs, and thereby to define at least the general nature of the Minoan language.…
Unsurprisingly, the assumption that the Linear B language was ‘not Greek’ got a rise out of the ever-cautious Bennett. In reply, Ventris asked him jokingly whether he was ‘secretly cooking up a Greek decipherment – out with it!’
His first technique was statistical, of the kind introduced earlier when discussing Bennett’s signary of Linear B. Using Bennett’s transcription of the tablets from Pylos, Ventris counted each sign group (word) with a distinct spelling, and discarded recurring sign groups. (This was necessary to avoid distortion of the statistics by counting multiple occurrences of common words, including names.) It left him with 5410 total occurrences of signs. He then counted the number of occurrences of each of 79 signs, taken from Bennett’s 89-sign list, and calculated their frequencies out of 1000. This enabled him to classify the 79signsinto three categories: Frequent (15 signs), Average (26 signs) and Infrequent (38 signs). In addition, he classified each sign according to its position in sign groups, such as initial, final, second or all positions. Overleaf is the sign frequency table he produced in Work Note 8. Note that the three signs , , and – which include the ‘double-axe’ and the ‘throne-and-sceptre’ – have high frequencies in the initial position. This suggested to Ventris that they were probably pure vowels. The reason is that in a syllabary of the CV (‘open’) type, the sign for a pure vowel will occur mainly at the beginning of a word, because in every other position a vowel will normally be subsumed into the syllabic sign. Thus, in English, ‘operator’ would be spelt syllabically o-pe-ra-to-r(o) and ‘anagram’ a-na-g(a)-ra-m(a), with no pure vowels except in initial position. A pure vowel would appear within a word only in examples such as ‘initial’, i-ni-ti-a-l(a) or ‘anionic’, a-ni-o-ni-c(i). The latter type of word turns out to be comparatively infrequent, whatever the language under consideration. (Fifteen minutes spent scanning an English dictionary will prove the point.)
Sign frequency table from Work Note 8, May 1951.
Other useful frequency analysis concerned sign combinations, e.g. two signs adjacent to one another in a sign group, since each language is characterized by certain frequent combinations of signs (such as ‘th’, ‘io’ or ‘qu’ in English) and infrequent combinations (such as ‘dh’, ‘ao’ or ‘qi’). But as Ventris admitted, frequency analysis has major limitations as a decipherment tool. He told Myres: ‘You are…quite right to be sceptical of the value of statistical studies as a direct key to decipherment: but I feel they are necessary to sketch in the background of fact against which any proposed decipherment will have to be tested.’
Another sign (‘button’) was found to occur with ‘average’ frequency but apparently exclusively in the final position of sign groups, as in the following tablet:
Tablet drawing by MV from Antiquity, 1953.
That the ‘button’ sign was a suffix and not part of the root was clear from the fact that there occurred pairs of sign groups of the following kind:
This suggested to Ventris that the ‘button’ might be a conjunction such as ‘and’, formed not in the English way as a separate word but as a syllable tacked on to the preceding word, like ‘-te’ in ancient Greek and ‘-que’ in the Latin phrase ‘Senatus Populusque Romanus’ (SPQR), meaning ‘The Senate and People of Rome’. In English, ‘-ly’ is tacked on to adjectives to form adverbs in the same way (e.g. quick-ly).
However, in a lengthy discussion of in Work Note 10, Ventris was cautious of jumping to the conclusions that, first, all occurrences of the ‘button’ sign meant ‘and’ and, second, the sign was the only way to write ‘and’ in Linear B. For he had noted certain final-position occurrences in sign groups where the ‘button’ sign did not appear to mean ‘and’ from its context (indeed the sign would later turn up functioning as an ordinary syllabic sign at the beginning of some Pylos sign groups). Also, he knew that Latin, for example, has the very common discrete conjunction ‘et’ meaning ‘and’, as well as the tacked-on ‘-que’. Lastly, he had noticed sign groups in the Pylos tablets which were apparently linked in sense by ‘and’ but where there was no ‘button’ sign written (as in English we write a list such as ‘London, Paris, Amsterdam and Berlin’ without writing ‘and’ between the first three cities).
Then there was the technique of scribal variation – the subject of Work Note 9 (written, somewhat confusingly, after Work Note 10 on ‘Minoan’ conjunctions). In other words, a search for variant spellings of the same word, such as ‘recognise/recognize’, ‘encyclopaedia/encyclopedia’ in English. Ventris looked for pairs of Linear B sign groups which appeared in two slightly different spellings, for example:
and he deduced that there was a close relationship between a sign in one group and the sign that had replaced it in the second group (perhaps a shared vowel but different consonant). For this technique to work, the two words have to be semantically the same and differ only graphically (by one or perhaps two signs), e.g. center/centre, sceptical/skeptical, inflection/inflexion, traveler/traveller, jewelry/jewellery, and must not be two words of completely different meaning, such as butter/butler, yellow/bellow, friction/fraction, prescribe/proscribe, which happen to have one sign different but are unrelated in their meaning. The obvious problem is how to establish this when one cannotreadthelanguage. The most promising examples in Linear B came from pairs of long words, or from scribal erasures of spelling mistakes in which it sometimes proved possible to read both the intended sign and the old, erased sign beneath it.
The existence of such 3500-year-old clay-tablet ‘rubbings out’ alerted Ventris to the possibility of uncorrected spelling mistakes in the tablets – which most certainly occur in cuneiform tablets and in Egyptian hieroglyphic inscriptions. ‘Since we are still so far from a full understanding of the Pylos inscriptions, it is perhaps rather presumptuous to point to any spellings in the sign group index as actual mistakes on the part of the Minoan scribe. Nevertheless there are a number of spellings which must look anomalous even to us’, Ventris noted. And he turned out to be right in this assumption: some Minoan scribes were more careful writers, or at least better spellers, than others.
But the most far-reaching technique used by Ventris – which he applied in the later Work Notes from the early summer of 1951 onwards, first to the Pylos tablets and then from April 1952 to the Knossos tablets (published in Scripta Minoa, volume 2) – was to search the sign groups for patterns of inflection. Evans had doodled this idea in the 1930s; Kober had drawn the inflectional ground plan in the 1940s; Ventris would design the elevations and build the structure in the 1950s.
In mid-June 1951, on this topic he wrote: ‘It is
possible to disagree here and there on whether a particular pair of sign groups really are inflected forms of the same word; but the basic existence of inflection, hinted at by Evans and confirmed with rather ponderous logic by Kober, is now beyond doubt. I must admit though, that at the time of Kober’s articles [i.e. up to 1948] on the subject I was inclined to follow Myres in dismissing most of her evidence for inflection as being merely alternative name endings’ – something like, to take a not-so-satisfactory English parallel, Parkins/Parkinson, Robins/Robinson, in which the name change could give the misleading appearance of being an inflection formed by adding ‘-on’ to the root. (This distinction – between inflection and alternative name endings – would prove highly relevant at the time of breakthrough in the decipherment.) By late August, he was willing to be more definite still: ‘the large and very interesting body of evidence which can be extracted from [the Pylos sign groups] makes the presence of inflection, first seriously demonstrated by Kober, a certainty.’ In due course, Ventris would be prepared to give the late American scholar further credit for her work on inflection, but he would never be warm in his praise. And given the consistent cold shoulder he had received from her throughout, even before her illness, one does not feel inclined to criticize him. (Champollion was far less generous to Young in the Egyptian decipherment.)
In Work Note 11, Ventris studied the question of inflection for gender. This is characteristic of common nouns in some Indo-European languages such as ancient Greek (e.g. doulos/doulē, ‘slave’), Latin (dominus/domina, ‘master/mistress’) and Italian (bambino/bambina, ‘child’), but not Danish or English; although English does use a few forms like ‘prince/princess’, the vast majority of common nouns do not vary with gender. Such gender distinction is also found in some non-Indo-European languages, though not in Etruscan, so far as we know from our very limited grasp of the language (Etruscan proper names take a feminine form, but common nouns do not).
From the evidence of the Linear B tablets, it looked to Ventris as if the language did mark masculine and feminine forms – as seen in the following pairs of sign groups, in which the discrimination between the masculine and feminine forms is based on the adjacency of these sign groups to masculine and feminine pictograms/logograms (see page 25):
These comparisons were of course grist to the grid, since they suggested that the final sign in each pair of sign groups shared either the same consonant (or conceivably the same vowel). From their contexts on the tablets, the first three pairs appeared to mean: ‘child’ (boy/girl); ‘total’ (male/female) – this guess had already been made by Evans and settled by Kober, as we know; and thirdly, ‘slave’ (male/female).
Inflection based on number, in other words singular and plural forms, was useful too. Where there were numerals on the tablets, singular and plural forms could often be distinguished; a numeral 1 obviously implied a singular form for the adjacent sign group. In English, noun plurals are generally inflected with the addition of an ‘-s’ or ‘-es’, or perhaps a change of syllable (e.g. woman/women) or even the addition of a syllable (child/children) – but there are some nouns that do not change (e.g. sheep, species). In Linear B, some sign groups seemed to be the same in both the singular and the plural, i.e. they did not inflect, while others showed the addition of an extra syllable (highlighted):
(The first four pairs were encountered above in relation to gender inflection.) Ventris observed that the plural inflection in the second group resembled the formation of plurals in Greek, but the lack of change shown by the first four plurals was surprising.
Thirdly, Ventris searched for patterns of inflection by case in what appeared to be nouns, following on directly from Kober and her ‘triplets’. But while Kober’s ‘triplets’ had been taken from the few tablets that had been published in 1946 (those from Knossos), Ventris was able to retrieve many more of such patterns from an exhaustive search of the Pylos sign groups, always bearing in mind, as Kober and Bennett had, that the context of the sign groups he was comparing must be similar. According to him, certain sign groups that he judged from their context to spell men’s names inflected in three ‘cases’ – nominative, genitive and dative (here called ‘prepositional’ because of some uncertainty) – but they followed at least six different types of declension. (Six may seem unduly complicated, but there are three noun declensions in ancient Greek and five in Latin.) Among the examples of this he reported in Work Note 14, in late August 1951, were the following:
Again, it was possible to guess that the syllabic signs representing the case suffixes, added to the root in each declension, would be likely to share a vowel or consonant; but caution was required, given the complexity of the declensions.
At this point, with Work Note 15, Ventris produced a second grid. Compiled in Athens and dated 28 September 1951 (he was then on his first visit to Greece, including Knossos), it is a big improvement on the first grid. Not only are more signs placed on it, they are more accurately positioned (judging of course with post-decipherment hindsight). But what is perhaps most striking is Ventris’s cautious avoidance of phonetic values, as compared to the first grid. He shows none. Clearly, his intensive study of the Pylos tablets – indicated by his notes at the head of the columns of the grid – had convinced him that he should try to avoid some of his earlier shaky procedures, such as allotting possible phonetic values on the basis of Etruscan analogies.
Ventris’s second syllabic grid, from Work Note 15, September 1951.
He commented: ‘the evidence which [the grid] offers now for phonetic values is not complete in itself,’ i.e. it was too soon for the substitution phase of the decipherment mentioned earlier. Then he outlined his view of future work: more analysis of the case endings of proper names; comparison of sign frequencies and sign combination frequencies with syllabic frequencies in known languages; study of the connections between the Cypriot syllabary and Linear A/B; comparison of Linear A with Linear B; and most important of all, study of the Knossos tablets, as yet unpublished by Myres. But he admitted that it was ‘doubtful’ that phonetic values could be derived from these methods; at some point, it would be necessary to substitute at least some phonetic values derived from a known language chosen because it resembled the ‘inflectional endings, vocabulary words and proper names’ of ‘Minoan’. For as he said, ‘to wait for a bilingual to help us to solve our problem is to cry for the moon.’ There was no Linear B equivalent of the Rosetta stone, and no practical prospect of finding one.
Deeply fascinated as he was by the problem, even Ventris was feeling the strain – not least because he had found and purchased a plot of land in Hampstead where he wanted to build the family house he and his wife had been discussing for almost ten years. He told Bennett in early September: ‘this summer’s work has rather upset my home life, and I’d like to ration my Minoan…during the next twelve-month.’ In Work Note 15, he chided everyone on the circulation list (now 31 scholars): ‘At the time of our joint Bulletin of 1950 [the Mid-Century Report], I had in mind to give up active Minoan studies owing to pressure of other work. I am now afraid that the considerable volume of these “Work Notes” may seem ironically to give the lie to this protestation, and that I may seem to have bitten off a more greedy chunk of the research than I can chew.… However, the other work still presses, and I should be very glad to see these studies taken further by others with whose methods I [am] in sympathy. I must put it on record, though, that at this writing only our colleagues in the United States [the now-dead Kober, and Bennett] seem to me to have so far shown the realistic approach. It is up to us in Europe to redress the balance.’
Strange though it may seem, at no time in his life would Ventris regard the decipherment of Linear B as a competitive race. Acting in the spirit of ‘group working’, his Work Notes and personal letters to scholars prove that he was always genuinely pleased to see others make progress in the field. Sometimes, one even has the sense that he would prefer not to be the one who would finally ‘crack’ the cod
e. All the time he was obsessed with the subject, some part of him seems to have felt that he should be pursuing something else: the neglected architectural work for which he had been trained.
Nevertheless, there were two more mailings of extensive Work Notes before the end of 1951. Then Ventris and his wife left London for a three-week skiing holiday in Europe. But by the end of January 1952, he was back again in the Highpoint flat, hard at work reviewing the Pylos tablets and preparing for the arrival within a couple of months of Scripta Minoa, volume 2.
Work Note 17, dated 15 February 1952, contains a third, and final, grid. This time, phonetic values, for both vowels and consonants, have reappeared. How did Ventris arrive at them? According to him, ‘The tentative phonetic values shown in the first column [i.e. the consonants] are not to be taken too seriously. They pay scarcely any regard to supposed similarities with the Cypriot syllabary, but represent the values which seem the most useful in giving an “Etruscoid” character to the Pylos names, words and inflections. This is still at the moment a very experimental procedure, but I am optimistic that before long it will prove the key to a reliable transliteration of the syllabary.’
Ventris’s third and final grid, from Work Note 17, February 1952.
In fact, his first guesses for the vocalic values in the vowel columns, i, o, e and a – are predominantly correct, being based partly on pure vowels’ high frequency of occurrence as initial signs; while his guesses for the consonants are mainly wrong, deriving as they do from his tenaciously held, but incorrect, theory that ‘Minoan’ was related to Etruscan. Furthermore, the few syllabic signs which are correct, such as C3V5 () = pa and C6V2 () = to, may have been influenced, consciously or unconsciously, by the resemblance of these two signs to the Cypriot signs for pa and to, despite Ventris’s disclaimer. (See the Cypriot syllabary on page 35.) All in all, the grid was a mixture of logic and intuition: a useful experimental tool, but far from infallible. It was emphatically not the unique technique that would unlock Linear B.
The Man Who Deciphered Linear B Page 8