Sunday, January 31, 2010

Cracking the libation formula - Part II

Reading back my previous posts, I realised that continuation was mentioned many times. So this time I fulfill one of my earlier promises, and continue the discussion of the famous Linear A libation formulae. (This post took long enough to complete as I had too little time to devote to the task last week.)

In the preceding post, we discussed a small but important part of the Linear A religious texts: Yet - literally - we did not get past the first word. In this chapter, we will attempt to analyse the deep structure of the whole Libation Formula, hoping to get insight into the meaning of its words. We shall make good use of the works of John Younger, who transliterated, classified and compared these inscriptions, and also made them available to the public on his website.

Our method is going to be a purely comparative one: using as little as possible input from other languages (not even Etruscan-Lemnian structures), our attempt will be a 'universal' one (informatics would call this 'platform-independent'). We only make one simple assumption: that the different texts found on libation vessels represent essentially the same grammatical structure, the same key words with slight variations, determined by the context (such as the subjects' number, the objects' number, and so on). Then, by comparing different sentences, looking at the word-formation carefully, we can attempt to build up a dependence-tree from the different words. And from these relations, with some luck one can assign the words their roles as subject, object, predicate or epithet.

So let us take our 'sample sentence' once again:
(α)A-TA-I-*301-WA-JA • (β)JA-DI-KI-TE-TE-DU-PU2-RE • (γ)JA-SA-SA-RA-ME • (δ)U-NA-KA-NA-SI • (ε)I-PI-NA-MA • (ζ)SI-RU-TE

As I mentioned it before, this is in fact a hybrid between two more-or-less complete inscriptions (PK Za 8 and IO Za 2). I made it up partly because the fragmentary nature of some finds would make it difficult to see the entire structure in one piece; and I also wanted this base form to feature the word JA-DI-KI-KI-TE-TE-DU-PU2-RE. (You will see what my motives were very soon.) Since this inscription above (with the constant variation of words in place β, but keeping all the other words α-ζ constant) represent the most common form, we will term this case as type #0 (or the base type).

As for the variations, inscriptions PK Za 11 and PK Za 12 shall represent type #1 and type #2, respectively. PK Za 12 (variant type #1) is depicted below:
(α)A-TA-I-*301-WA-JA • (β)A-DI-TE-TE-? (•) SI-? • (γ)?-RA-ME (•) A-?-NE • (δ)U-NA-RU-KA-?-JA-SI • (η)A-PA-DU-PA-?-JA (•) A-? (•) (θ)?-JA-PA-QA

As we see, word ε of the base formula was substituted with a chain of words I labelled η-θ. These do not occur anywhere outside these two Palaikastro libation vessels, so we cannot analyse them meaningfully with our comparisons. Still, we see many remarkable things: the words noticeably changed in position δ, and less profoundly but recognisably in position β. The inscription PK Za 11 (variant type #2) shows even more profound changes:
(α)A-TA-I-*301-WA-E • (β)A-DI-KI-TE-TE-?-DA • PI-TE-RI • A-KO-A-NE • (γ)A-SA-SA-RA-ME • (δ)U-NA-RU-KA-NA-TI • (ε)I-PI-NA-MI-NA • (ζ)SI-RU-? (•) (θ)I-NA-JA-PA-QA

Here barely a few words kept its original form. In addition to changes seen previously in positions β and δ, further important changes occurred at position α, γ, δ and ε. Pure common sense suggests that these secondary changes are independent from those already seen on PK Za 12. Since we now have a chain of words in position β instead of a single one, it is tempting to assume that we have some kind of singular-to-plural change in the structure. But let us not stop here, and take a look at the next more-or-less clearly identifiable class:

Variation classes #3 and #4 consist of inscriptions KN Za 10 and PR Za 1. These texts display a different type of deviation from our case #0. Most obviously, these begin with TA-N-, and - consequently - are much shorter, lacking parts δ-ζ:
(α)TA-NU-MU-TI • (γ)JA-SA-SA-RA-MA-NA • (β)DA-WA-? (•) DU-WA-TO (•) (η)I-JA-?

(α)TA-NA-SU-TE-?-KE • (β)SE-TO-I-JA • (γ)A-SA-SA-RA-ME

Here the mark η stands for unique parts not repeated on any other document. The truly interesting feature of these inscriptions (that otherwise have little in common) are the semi-regular changes in words α and γ. In the first example, we see a genuine case of declensional change on word γ (the addition of an intusive suffix -A-NA), concomittant with a different case-ending of the word at position α . Case #4 shows a different type of change on the same words: this time something reminiscent of the changes seen on PK Za 11 (case #2). The occurrance of an -E ending on word α is followed by the removal of the initial J- from word γ. Yet these two cases are too different to establish something systematic out of them. The only conclusion we can make is that "whenever word α changes its ending, corresponding changes in word γ will follow".

Finally, for comparison, we will inspect yet another class of the libation formula. These are not true grammatical variants, but instead represent the case when some words of the formula are substituted with logograms. Some pithoi contain such inscriptions, like SY Za 2 and ZA Zb 3:
(α)A-TA-I-*301-WA-JA • (β)JA-SU-MA-TU • (γ?)olives • A-JA • (δ)U-NA-KA-NA-SI • (ε?)olive-oil

(α)A-TA-I-*301-DE-KA • (β)A-RE-PI-RE-NA • TI-TI-KU • (γ?)wine • 32 • (η)DI-DI-KA-SE • A-SA-MU-NE • A-SE

It is quite interesting to see that positions γ and ε are used by names for commodities in these cases. It reminds us that whatever JA-SA-SA-RA-ME and I-PI-NA-MA stood for, had likely more to do with the fluids used for the libation than any other thing. But again, we have no direct proof, and the fact that JA-SA-SA-RA-ME in fact occurs on other objects such as statues as well, undermines any attempt to translate it as "olives", ''olive-oil' or 'wine'.

What could we have learnt from all the above examples ? If we gather all our observations on words showing corresponding changes, we can get the table below:

case # α β γ δ ε others
case 0 A-T-α-JA J-β J-γ δ-SI ε (misc)
case 1 A-T-α-JA β J-γ(?) δ(RU)-SI ε (misc)
case 2 A-T-α-E β γ δ(RU)-TI ε-I-NA (misc)
case 3 TA-N-α-TI β J-γ-A-NA missing missing (misc)
case 4 TA-N-α-E β γ missing missing (misc)

We can also use the (intuitive) assumption that additional changes in class #2 with respect to class #1 can be separated. With this, we can get a greatly simplified set of relations. The following rules can be obtained:

Rule I:
word β lacking J- ↔ multiple words at β
↔ word δ containing -RU- infix
Rule II: word α ending -E ↔ word ε ending -I-NA ↔
↔ word δ ending -A-TI
Rule III: word α ending -TI ↔ word γ ending -A-NA
Rule IV: word α ending -E ↔ word γ lacking J-

It is interesting to look at the initial J- elements. This particle has been variously interpreted from 'grammatical prefix' to 'dialectal or spelling variant'. But neither of these interpretations appears to be fully correct. Since we have specimens of the libation formula from Palaikastro with J- and without J-, the difference is anything but dialectal. However, we also know that the initial J- is a minor element of the language at best - prefixes are unlikely to be part of the Minoan language. This leaves us only one option: like the initial TA-N- (*tan), the demonstrative pronoun's accusative case, the initial I- or J- should be a pronoun or article (*i), simply written together with the word it refers to - as usual in Linear B.

No matter what this I- or J- exactly means, its role is secondary, and does not really modify the grammatical case of the word. But it still turned out to be pretty useful to detect declensional dependence relations. We can also depict these relations in the form of a 'dependence-tree' graph, as shown below:

Looking at this graph, there are two words that occupy a central position: α and δ. So either of them must be the predicate, hiding a verb or verbal derivative. We cannot decide this without getting some help from the previously-deciphered phrases. For example it is clear that word β is most commonly a place-name, and thus it has something to do with the subject of the sentence. On the other hand, word α appears to contain some sort of demonstrative pronominal particle (see part I of this post). Therefore its chances of being the predicate are low.

From here, a fairly trivial conclusion would be to regard word δ as the predicate - as an active verb - and word α as its subject. Despite referring to places and persons, word β cannot be the subject, due to formation issues in word δ (a verb - in a suffixing language - should not carry the plural marker buried inside the stem). But word δ still modifies the predicate, so we must assume the predicate refers to the object offered (word α) modified with its origin (word β) Words like γ, ε modify the subject, while word ζ does not seem to modify anything.

As for word δ, it not only occupies an important position in the structure, but also has a high number of forms possible. It is important to observe that not only the changes of the subject (word α, that also determines the form of both γ and ε) are reflected in the predicate δ, but also the changes of part β - so δ might be a compound word, like 'to give libation' (or even 'pouring of libation', if its not a verb). For example if U-NA-KA-NA-SI is "gives libation", then U-NA-RU-KA-NI-JA-SI (variant of U-NA-RU-KA-NA-SI) could mean "gives libations" and U-NA-RU-KA-NA-TI "give libations". The lack of J- on part β coincides with the presence of additional words in this position, therefore a singular-to-plural change is not a bad assumption at β, and that would naturally change the first part of verb δ (as multiple donors are expected to perform multiple libations).

The endings of JA-SA-SA-RA-ME and I-PI-NA-MA change in a distinct, though fairly similar way. In addition, they are both determined by the putative subject α. These two facts suggest that these two words fulfill a similar grammatical role in the sentence, further specifying the object of the action of libation. Finally, word ζ (SI-RU-TE) seems to incur no changes at all throughout the formulae, and thus might be something like an adverb. A suggested (though heavily tentative) translation of the entire formula is depicted below (I used Miguel Valerio's concept of translating JA-DI-KI-TE-TE-DU-PU2-RE as the 'Diktaian Master'. Other words were haphazardly given a meaning, except I-PI-NA-MA - I will devote an entire post to that later):

There is one more little bit of trouble with the above-mentioned interpretation: namely, the fact that texts beginning with TA-NA- are always shorter than the A-TA- types and these ones never contain the phrase 'U-NA-KA-NA-SI'. Since this class is numerous, one cannot help, but imagine that in these cases, the word α takes the role of the predicate. These cases feature a different composition of word α, and that suggests that its original noun was supplanted by or supplemented with a participle or an active verb. For example: "this vessel (nom)" contrasted to "this vessel (acc) gave" (Special thanks go to Glen Gordon for pointing out that TA-NA- is accusative case). The reconstruction featured on the figure above was based on the latter structure.


  1. Interesting blog.
    I've been working on several ill-understood languages as well. I think they are Hurrian dialects.
    I would personally recommend to check your analyses against Hurrian.


  2. I must confess I did play with the thought many years ago, when I first became interested in the Linear A inscriptions. It was too bad I could not encounter any Hurran glossary large enough to serve as a meaningful basis of comparison. Initially I thought that the word-initial J- came from some Hurrian pronoun, but later rejected the idea. Since then I have taken the position that Minoan is much closer related to Cypriot, Etruscan and Lemnian than any other surrounding language. But what you say is certainly worth a try - to see if we can come to some good grammatical parallels and well-fitting words.

  3. I really don't see why people continue to push a desperate connection between Minoan and Hurrian of all things. If we just look at it geographically, we see that the Hurro-Urartian group gravitates around the region of Lake Van in *eastern* Turkey. I wouldn't be averse to the idea however that some Hurrian words made their way into Minoan via Hittite or Luwian. That at least is plausible but exceedingly hard to prove at the present time given such little information.

    As for your analysis, Andras, well done although you could cut it down a few paragraphs. ;o) The phrase JA-DI-KI-TE-TE-DU-PU2-RE is surely two words: Adíkitete dupure. To an Etruscan, it would look like a locative noun phrase (ie. Minoan -e-te = Etruscan -e-θi []+'in'). It's already surmised by many that Adíkitu means 'Mount Dikte'.

  4. I made a Hurrian vocabulary available here:
    There are plenty of indications that Eteo-cypriot and Carian are also very close to Hurrian. And I think it is possible to decipher them with Hurrian.

    I'm not sure what the connection between Minoan and Hurrian really is, but in all cases Hurrian, Carian and Eteo-Cypriot have about nothing in common with Etruscan.
    And I can't see any reason to make Etruscan a non native and non autochthonous language.

    I'm neutral about Linear A language.
    Eteo-Cretan as written in Greek alphabet is completely un-hurrian sounding.
    And I can't see what Etruscan would bring here.


  5. Thank you Glen, for adding one small but important detail: the explanation of JA-DI-KI-TE-TE-DU-PU2-RE. (I must admit, I was so busy explaining the other words on the libation tables I totally forgot about this one).

    I also believe that the interpretation of the -TE (perhaps *-ethe) endings as a locative is essentially correct. The only thing I could not explain up to date is the fine distinction between the -TE and -TI endings (the latter, too, has been implicated as a locative).

  6. Dear Yangg,

    Thank you for providing this resource.

    As for the Etruscan connection, it is not self-understanding, nor intuitively trivial. I came to join this camp, because of the pronominal and declensional system found in Minoan Linear A and Cypriote Linear C shows undeniable similarity to Etruscan. The Cypriot and Cretan languages appear to be heavily related to each other: take the example of the TA-N- particles (*tan = accusative case of the demonstrative pronoun). If we work on endings, we may come to similar conclusions: the vowel-consonant-vowel type suffixes of Minoan words appear quite different from the Indo-European system but have possible parallels in Etruscan. The semitic interpretations, on the other hand, are often absurd.

    Still, there are still some words, like PO-TO-KU-RO that suggest IE connections. How can we reconcile that? Because there is an increasing number of scholars who believe the Etruscan, Rhaetian and Lemnian languages actually orginate in the Aegean, and are distantly related to Proto-Indo-European, there may not be any contradictions at all. Once I read an intrepretation of Eteo-Cypriot texts that used the Aegean theory (these were the first meaningful translations I have ever seen), it convinced me immediately that it is more than worthwhile to search for such parallels.

    The Aegean theory is important because it can explain a baffling problem: both Cretan, Cypriot and mainland Anatolian languages of the 2nd millenium B.C. seem to lack voiced consonants, and disfavour or straightforwardly lack vowels like E or O. In turn, they may have aspirated consonants (TH, PH, etc.). Etruscan, Rhaetian and Lemnian share exactly the same features. But the Anatolian languages like Luwian or Hittite are Indo-European, and this development was clearly shown to be secondary in the Anatolian branch (i.e. de-voicing of voiced consonants). The best possible explanation is a Bronze-age areal effect between nearby languages (that is, the Aegean and related languages influenced the Anatolian ones or an unknown third language influenced both). This areal effect also extends to Lemnian. On the other hand, Hurrian seems unaffected.

    As for Hurrian, I would love to see a consistent theory that explains its connections with Indo-European languages. Yet I would not be surprised if the Hurrian-Urartuian family would turn out to have common roots with some Caucasian languages instead.

  7. Correction to my post above:

    Hittite did not lack voiced consonants: nevertheless, it is believed to have undergone a consonantal shift in its stops according to some scholars:
    voiced -> voiceless; voiceless -> aspirated; aspirated voiced -> voiced w/o aspiration
    (i.e. D->T; T->TH [-tt- in script]; DH -> D). This Anatolian phonological theory is no way a universally accepted one, yet interesting explanation of hittite consonantal gemination.

  8. How do you conclude that Aegean languages "disfavour or straightforwardly lack vowels like E or O"? What reason is there against *e and *o reconstructed for Proto-Aegean as well. (Afterall, vee-shaped 5-vowels sytems are very common the world over.)

  9. As for the E and O sounds, their presence in Minoan was always (how do I put this) problematic. This is because of several factors: One is the absolute rarity of these sounds. This is not a subjective 'rarity': I did some statistical calculations on the frequency of signs (of course, to help the decipherment), and that did give some interesting results: For example, the frequency of 'O' sounds (judged by the frequency of the corresponding syllabograms) is only about 5% [and this is mostly from sign RO]. The same time 'A' has about 35%, 'U' and 'I' 20-25% frequency. Ther vowel 'E' has another interesting feature: Despite its moderate overall frequency (15%), it is only found in about 1-2% of the word-initial vowels (half of the initial vowels is 'A', and - while still disfavoured - 'O' does not face the same level of discrimination in initial positions as 'E').

    The second argument (that probably comes from the first, i.e. the low overall frequency), is the extreme rarity or total lack of certain 'O'-based signs. Take the example of 'SO': it occurs about 2 times (in the same word) in all the corpus. Sign 'QO' does not occur at all. 'NO' is also extremely hard to find (perhaps the *28b sign IS 'NO', but that only occurs 3 times). There is also another problem with the Minoan 'O': namely, ther are sign-groups, where 'O' and 'U' seem exchangeable to some extent (apart from the difference in junction): take the example of SU-PU2-WA and A-SI SU-PO-A.

    For the last, the value of certain signs hints at their origins, and that does tell something about the original value of 'O' signs: Take the example of Lin A *79 = DO. It depicts an eye. I have a feeling that it does have something to do with Luwian Dawa = 'eye' and Etruscan Tva = 'to see'. In this case, it might have arisen through the path *tawa -> *tau -> *to [DO]. While I do not deny the existence of 'O' and 'E' in Minoan (some scholars do), they are still pretty rare compared to other sounds. Just do a similar statistics for Greek, and you will see what I did mean.

  10. Bayndor [sounds like Gold Bath in French!] wrote : hittite consonantal gemination

    I think you are confusing the features of the inadequate cuneiform system, which does not have a clear and straightforward way of writing voiced consonants with the features of the real language(s).
    It's nonsense to posit REAL geminates. these are just graphic geminates.


  11. As for Hurrian and PIE connections, I've written a book draft on that topic with Allan Bomhard.
    I definitely think that Hurrian and PIE are close relatives. Hurrian is very close to the kind of PIE hypothesized by some people like Lehmann and alias.
    This is not exclusive of possible connections with Caucasic languages as well. I'm not in a kind of "it's X or Y" approach. I tend to think that both are true.


  12. Naturally, Yangg's delusional Hurrian-PIE connections are beneath intelligent debate.

    "Yangg" also goes by "Arnaud Fournet" and trolls Cybalist... incessantly.

  13. Maybe the issue is your own inability with having an intelligent debate.
    By the way, my study of PIE-Hurrian connections is published here: