Treatment of consonantal clusters in Linear A and B

Linear A and B are - in some way - very imperfect writing systems. This feature almost immediately reveals itself, once we attempt to transcribe some (let's say, English) words into it. For example: "Cargo" would become KA-KO, "Sleepy" would be RI-PI and "Window" WI-DO. Apart from the simple absence of certain consonants (such as G or the distinction between R and L), there is a much graver problem: this script has a seriuous trouble when it comes to writing down consonantal clusters (that is, more than one consonant following eachother without vowel).

What we have seen in the previous (horrifying) examples, is the very thing that made (and makes) decipherment of linear B documents hard. Since the underlying language is not one like the Japansese - that has virtually no consonantal clusters. And to write down the Greek words, the scribes had to make up some methods, that included ommission of certain consonants. That's how the horrendous examples shown above were born: simply by following the same rules.

What were these rules? And were there any rules indeed? One would wonder if there were some rules for ommission, not some haphazard "scribe's choice" ones, it may help us reconstituting the original words. Even if we may never know the exact consonant value (that has been lost), we may have a good guess at it. And - if we make a small analysis like the one below - we may ascertain, that indeed, there were rules to write consonantal clusters with the Minoan script!

Theoretically, if two consonants meet eachother in a word, and our script has no pure consonantal signs, we can do two things: First, we may insert a vowel fitting into the gap, thereby "resolving" the cluster. This is what current-day Japanese writers do, when encountering a foreign word. But there is another way to work around this problem: One may simply ommit a member (intuitively, it is the easiest to do this with the first one) of the cluster, thereby "simplifying" it. Minoan (any Mycenean) scribes used both methods in combination with each other, as the table below shows:

		FIRST	CONSONANT
SECOND CONSO- NANT	Nasals (M+N)	Bilabial plosives  (B+P)	Alveo-dental  plosives   (D+T)	Velar plosives  (G+K)	Alveolar fricatives  (Z+S)	Trill / lat. approx.    (R+L)
(M+N)	Resolved	Resolved	Resolved	Resolved	Simplified	Simplified
(B+P)	Simplified	N/A	(?)	(?)	Simplified	Simplified
(D+T)	Simplified	Resolved	N/A	Resolved	Simplified	Simplified
(G+K)	Simplified	(?)	(?)	N/A	Simplified	Simplified
(Z+S)	Simplified	(?)	(?)	Resolved	N/A	(?)
(R+L)	Simplified	Resolved	Resolved	Resolved	(?)	N/A

(This table was generated from the representative dataset of Lin B words shown in the Mycenan Greek "dictionary" of Markos Gavalas. The ? signs shows clusters not found in any of the words, while some consonantal clusters on the diagonal, marked by N/A are not expected to occur at all.)

From the above, one may easily formulate the rule: There are two types of consonant according to this script, "strong" and "weak" ones. "Strong consonants" (plosives) keep their value when written down, and therefore any cluster beginning with a plosive must be resolved by inserting a vowel. "Weak consonants" (nasals, fricatives, trills or lateral approximants) on the other hand, are only written down if that cluster begins with a "strong" consonant. Otherwise, if they collide with a strong consonant, they are simply ommitted. The only exception from this "strong-weak" rule is the case when two "weak" consonants collide. In this case, ommission (and thus simplification) occurs in most cases, except the case of two nasals, when the cluster (e.g. -MN- like in Agamemnon) gets resolved with the insertion of a vowel.

Now, that we have the rule, we can go back to our inscriptions and use our knowledge to find (or at least guess) the original forms of the words. We can go back as far as the Linear A (which is at least several hundred years older than the Linear B script we set our rules on). But what guarantees that these writing rules were already used in that age? Of course, we have not enough "solid" evidence to state that with certainty (as we do not know the underlying language). The only thing we know, is that words of Minoan origin (in the Greek language) often contained such clusters (e.g. words with -nthos endings). So it is unlikely that the Minoan language was like the Japanese, lacking these problematic clusters. Yet we cannot find them in the Linear A script, meaning that they must have been either resolved or simplified.

For a possible example of resolution, we find words like "TA-NU-NI-KI-NA" (PL Zf 1.1). One may easily discover that the second part of this word is related to U-NA-KA-NA-SI. (The first part, TA-N- possibly means "The") Indeed, it is very likely that the second "I" in the fomer word was only insterted to resolve the cluster -KN-. On the same spot, the latter word has "A" - again, likely as a result of cluster resolution. This resolution is perfectly in-line with the rules we have seen above, in the case of Linear B.

For cluster simplification, we see other notable examples: let us consider the word "PA-I-TO"
(HT97 & HT120). This is a place-name, and - since it occurs in Linear B documents as well - we know that it stands for Phaistos, an important centre of Neo-palatial Minoan Crete. The mere fact that the same simplification rule was used to write its name in Linear A as Linear B (-ST- was simplified to -T-) strongly implies that the rule we discovered in Linear B already existed (in essentially the same form) in Linear A as well.

Cracking the libation formula - Part I

The so-called "libation formulae" are one of the most famous minoan inscriptions (well, apart from the Phaistos disc). A dozens of very similar texts were encountered on sacred objects (most typically libation-pouring vessels). With their length (at least 6 words = approximately two sentences) they constitute the longest known Linear A inscriptions. Sadly, there are not many of them recovered without illegible or broken off parts. But still, these are of an enormous value to those who try to delve deeper into Minoan grammar and vocabulary.

A sample inscription would look like the following:

A-TA-I-*301-WA-JA • JA-DI-KI-TE-TE-DU-PU2-RE • JA-SA-SA-RA-ME • U-NA-KA-NA-SI • I-PI-NA-MA • SI-RU-TE

(NOTE: in truth, this is a hybrid of PK Za15 and KO Za1 - to show a simplified example)

Unfortunately, the number of straightforwardly identifyable elements is close to nil. Yet, with more cautious examination of variations in formulae and comparison of the text with the vocabularies of other languages (such as Eteocypriot or Etruscan) a number of things can be made clearer.


This time, we shall look at the first word of this mysterious formula. This word usually consists of about 6 syllabary signs - already strikingly long, given that most expressions or names on the Linear A tablets have typically only 2-3 signs. But the structure of this word offers an immediate explanation, if we look at the different variants:

A-TA-I-*301-WA-JA (PK Za12)
A-TA-I-*301-WA-E (PK Za11)

TA-NA-I-*301-U-TI-NU (IO Za6)
TA-NA-RA-TE-U-TI-NU (IO Za2.2)
TA-NA-SU-TE-?-KE (PR Za1)

Here we can see that the different parts can be exchanged independently. In other terms, this is a compound word.

The first part of the compound seem to be either A-TA or TA-NA. But the separation A-T / TA-N seems equally if not more correct (see the explanation below). Now, the only question remaining: What on earth could this particle mean ?

If we look at the Eteo-Cypriot inscriptions, we might get a good hint for its meaning. It is very common for Eteocypriot inscriptions to contain the particle 'TAN-'. An example: WI-TI-LE RA-NU TA-NU MU-NO-TI A-I-LO. The final 'U' ending of TA-NU seems more of an insertion (to resolve the cluster -NM-), than a truely existing wovel. If we go back to Minoan Linear A, we can find similar examples with the same particle, e.g. : TA-NU-MU-TI (KN Za10).

There are many more-or-less substantiated tranlsations of the Eteo-Cypriot texts. One of the best (in my opinion) is based on the Etruscan-Lemnian vocabulary and grammmatics; this translation has been circulating around the net for quite some time (unfortunately, without the designation of its author). It was this translation attempt, that theoretized the existence of a demonstrative pronoun "TAN" = "This" in the Eteo-Cypriot language. It is a very interesting and suprising turn of events, that this translation almost perfectly fits the Linear A libation formula as well. The assumption that this long sentence began with the word "This..." is more than sensible.

Now we have gotten a suggested reading for TA-N. But what about the A-T particle it alternates with? Is it another related pronoun?
To answer the second question, the Eteocretan inscriptions provide another unexpected hint. These inscriptions recorded a non-greek language spoken in the classical Crete - widely theoretized to be a late descendant of Minoan. One of the Drerian inscriptions has the following reading: ET ISALABRE.....MEN INAI ISALURIA. The text is fortunately bilingual, the greek part speaks of offerings consisting of goat's cheese. In this context, the substitution of "ET" = "The" was already suggested by Van Effenterre et al. But what do we have here? A definite article that likely stemmed from some earlier Minoan word. The particle "AT" = "The" perfectly matches that in the Linear A Libation formula. This also reinfrces our previous theory about the meaning of "TAN". Thus the libation formula likely begins with an (attached) particle 'The....' or 'This....'!

The attachment of pronominal or demonstrative particles to the following word should not be surprising - scribes using the Linear B script did exactly the same with the Greek pronouns, fusing them with the following words, without any word-divider.

But let us go further in analysing this word. The next part in this compound appears to be most commonly I-*301 or rather (see the derivation above) A-I-*301. The major obstacle here is the missing phonetic value for Linear A *301 (the 'slave' or 'acrobat' sign). Since it is not an easy task to find the missing value (more on this in a later post), we have almost no clue of the meaning of this word. Unless we make a bold move and substitute a fitting Etruscan-Lemnian stem here. The best (given the context, and the rare *AI diphtongal cluster) appears to be the etruscan word-stem *AIS = "God, Divinity". If so, the value for *301 needs to be of either the S- or the Z-series (perhaps *ZU). Notwithstanding, the reading for a libation text beginning with 'This/The god...' appears incorrect. Therefore we have to assume that whatever A-I-*301-... meant, was more of a "divine gift" or "divine sacrifice" than being simply 'god' or 'gods'.

As for the last part (endings) of the formula, these were very differently interpeted by the scholars up to date. Some saw it as verbal endings (allowing the reading 'this - divine gift - was given'); another view would be to regard them as nominal(declensional) endings (admitting a reading 'this - divine offering-from/with/at/etc.') One thing is certain: This ending joins with the A-I-*301 particle with either W or U - making it suspicious that the (unknown) *301 sign ends with 'U'. At the same time, this joining suggests that the parts -WA-JA and -U-TI-NU form a true compound word with A-I-*301, and not just a random joining. The alternation of -WA-JA with the -WA-E ending coincides with the removal of the J- prosthesis from JA-SA-SA-RA-ME, and thus likely represents a putative mark of plural (verbal or nominal, but the latter seems more likely). We shall treat this phenomenon later on, once we analysed all the words in the inscription!

The Linear A *306 sign: searching for a value

Trying to assign the most signs of linear A to their linear B counterparts, one sooner or later stumbles into really problematic questions. After the assignment of the simpler signs (that were already done more than a decade ago, see the works of Jean-Pierre Olivier and Louis Godart), a handful of unique ones remain, that do not show obvious similarity to one or more of the (yet unassigned) linear B signs. Well, for at least for the first sight...

The LinA sign I will treat here bears the label of *306 - it is so high because scholars initially believed it to be a logogram. I would not say it is not used as a logogram on some places (e.g. KH6, where it likely stands for some kind of animal), but in most of its occurrances, it is a phonetic sign. Examples include the tablets HT6, HT119 and HT122 (*306-TU), HT115 (*306-TU-JA), HT122 (?-*306-KI-TA2)and HTwc3017 (DI-*306). But this sign is not particularly frequent, and these examples are - unfortunately - insufficient to guess at the phonetic value directly. All we can say is that the syllabe it represented was prone to stand word-initially.

If we believe (and this is clearly a hypothesis) that the use of this sign continued into the mycaenean era, we may attempt to search for a parallel in Linear B. As I stated before, sign LinA *306 belongs to the group of "animal-signs", i.e. those depicting various (domesticated) animals. Most of them have heavily simplified graphical structure, and are often used as logograms as well. Such examples of somewhat similar signs include the LinAB *23 sign (with the value 'MU'), that is sometimes used as a logogram for "cattle" or the LinAB *21 signs (phonetic value: QI), depicting (albeit simplified) the head of a sheep (and being a logogram for the latter). Fortunately, this group of signs is very limited in its numbers and their structure is very characteristic: with 1 line inserted below the main bulk ("head") as a "neck" and with an occasional addition of a "skirt" (an additinal vertical line) or "pants" (2 horizontal lines) to indicate the gender of animals.




If we look at the sign *306, we can guess the following: the animal it stood for is female (as *306 always bears "skirt"), and it has 2 prominent appendages on its head (horns or long ears), in addition to a large, bulky head. Now, if we go to the linear B signs, we may attempt to search for a sign with similar characters. However, one shall quickly find that the signs for "sheep" (LinAB *21), "goat" (LinAB *22), "horse" (LinB *105, LinA *336), "cattle" (LinAB *23) or "pigs" (LinAB *85) were already assigned to linear A signs, with great certainty. So the job of finding a counterpart is not that easy. Fortunately, we also know that LinA *306 also had a phonetic value used reasonably regularly. So we may search among the purely phonetic signs as well. There we may encounter LinB *42, that has a similar graphical structure to LinA *306. It has a skirt-like triangular lower part, with a line-like head, containing 2 appendages behind. So it is not beyond reason that linB *42 originally represented some kind of animal. Unfortunately, I cannot cite even a single unambiguous example, where linB *42 would have been used as a logogram. This leaves us a good guess at what kind of animal it may depict. My guess would go at "mule" or "donkey", (the latter maybe because the mycaenean greek word "Onos" = donkey rhymes well with the phonetic value of LinB *42 = WO ) but that is purely hypothetic.

As I mentioned before, the sign LinB*42 has the phonetic value 'WO' in linear B documents. But could a sign beginning with 'W' stand so frequently word-initially in Linear A? Without dismissing the hypothesis that LinA *306 is indeed 'WO' with no proposed phonetic change between LinA and LinB, let us look at other examples of 'W'-signs! A striking discovery is that both WA and WI (the only two such signs attested with certainty) are unexpectedly prone to stand word-initially. So, after all, words beginning with "WO-" in Minoan might not be misreadings at all.

Junction effects in Linear A script

The Linear B script is pretty heavily imperfect. This is caused by its simplicity (remember, Linear A or B had a comparable number of characters to the modern japanese syllabaries Hiragana and Katakana). All these systems represent only open syllabes. So consonantal clusters can only be written if one inserts auxiliary vowels between the consonants. Yet the mycaenean (and very likely the minoan) scribes did not bother themselves with doing this: they simply ommitted consonants (and often, vowels too) in case of encountering a cluster.

Thus for the consonants, the minoan Linear B is an imperfect system. This is not to say that there were no rules to tell whether to resolve or simplify a cluster (more on this in another post later). But if some clusters were consistently simplified, we may never tell from an extict word if it had a cluster or not. We know, for example that the place-name PA-I-TO (LinA and LinB) has to be spelled as Phaistos (at least in linB), but we would never be able to tell whether it had any 'S' in the middle if this name would be extinct. But just let us leave for the consonant problem for later. What is more interesting for us, to look at the representation of vowels in Linear A. Signs representing simple vowels in linear B are well-represented for all 5 vowels for linear A as well:

Signs with A, I, or U are most common, while those with E or O are much rarer. What really interesting is regarding the vowels, is the relative scarcity of diphtongs. Unlike linear B, where special signs (thought to represent diphtongs) are well-represented, these signs (most of them exist in linear A as well) occur very rarely in linear A. An even more interesting thing is the presence of 'junctional effects' when certain vowel-vowel signs meet each other. For example, if we have a meeting of U and A vowels, the minoan scribes have consistently rendered it to U-WA, perhaps corresponding to a lingustic tendency to avoid unwanted diphtongs. Such tendency cannot be observed in Linear B, where a meeting of vowels such as e.g. U-A was allowed without modification. If we represent all the possible junctions of vowels in Linear A, our table would look like as follows:

      A          O          E          I          U

A      A-A        O-A        E-JA       I-JA       U-WA

O      A-O        (?)       (E-JO?)    (I-JO?)    (U-WO?)

E      A-E        O-E        (?)        I-JE      (U-WE?)

I      A-I        O-I        E-I*       I-I*       U-WI

U      A-U       (O-U?)      E-JU       I-JU       U-U*

The junctions or signs shown with ? are not well attested. Apart from those, the tendency can be clearly seen: all junctions are allowed, though those that contain a semovowel as their first part (U=W, I=J) will have their following signs rendered to the corresponding CV-type signs. Signs ending in E behave as if they ended in I. An interesing case (*), and the only exception from the above rule is the case of homogenous joinings (I-I and U-U). In this case, the syllabes JI and WU will get rendered to simple I and U. The same happens for E-I (JI->I).

These vowel junctional effects can be used to effectively predict a (vowel) value of unknown or dubious signs. If an unknown sign is consistently followed by syllabes beginning with W+?, then it is highly probable that the sign had 'U' as a vowel. But this is never 100%, given that W+? signs can also stand alone (e.g. A-MA-WA-SI, the presence of WA does not imply the preceding sign to end with 'U'). Nevertheless, the same rule can be used to exclude given sign-values with certainty. If we observe an unknown sign to form a junction with A without the A-sign turning into WA or JA, then the unknown sign simply cannot end in U or I.

This exclusion rule can be applied to the Linear A sign *79 ('eye') to show why it cannot represent the value 'ZU' (that was assigned to it by John Younger et al.). The case-ruling example we find on the tablet ZA4,row a.5 where the term QE-SI-*79-E can be read. The same name recurs on tablet ZA15. Now, if the reading of *79 were 'ZU', we would rather expect an ending ZU-WE (with a not well characterised linear A WE sign) and NOT ZU-E. On the other hand, the value suggested and used by many (e.g Glen Gordon) for this linear A sign: 'DO', fits perfectly, as DO-E is absolutely possible.

But the case of LinA *79 has to approached with care. Apparently, there are two distinct LinearB signs (*79 and *14) corresponding to single cluster (*79) Linear A. I label it as a cluster, as it contains signs of very variable design: it is easily possible that there are two signs lumped into a single cluster: at least one of these is (with resonably high probability) is the Linear A counterpart of Linear B 'DO' sign (LinB *14).

Linear A signs with value 'PE' and 'RO2'

As the language of the Minoan Crete is irrevocably lost, the only clue to reconstruct bits of it, lies with the only written remnants of Crete's history from that age: the famous Linear A and Hieroglyphic documents (not counting the Phaistos disc). But with no one to tell us how to read those inscriptions, currently our only clue to the true phonetic values those signs hide, is some sort of 'back-tracing' of signs from the Linear B system.

I won't cover details of the Linear B writing system: one can find enough appropriate desription of the Linear B syllabary in any library or on the web. It is enough to say that the Linear B "alphabet" is one of the simpler writing systems of the ancient middle east: almost every (if not all) of its signs are of the structure CV or V (C=consonant, V=Vowel). This syllabary encompasses roughly about 100 signs (including those with an unidentified value). In addition to the syllabary, various ideograms are also used on the tablets as tools of a scribal stenography.

If one looks at the Linear A signs, similarities with the (younger) Linear B alphabet become too obvious to be ignored. Many of the signs can be identified one-to-one without doubt. But there are a handful of them, that do not display such counterparts. At least not one to be identified at first sight. But some of them have gradually found their counterpart thanks to the relentless work of scholars. Currently, there are so many identified linear A and B sign-pairs, that we may assume: almost all Linear B signs must have a Linear A ancestor, and most Linear A signs were continued to see use in the (Greek) Linear B inscriptions.

So, if we have a one-to-one relationship regarding most of these signs, then why not go and find the missing pairs? This is exactly what I tried with the following Linear B signs: LinB *72 (PE) and LinB *68 (RO2). Up to date, there have been but little suggestions on whether these signs had any Linear A counterpart.

As for the sign LinB *72, the sign displays a characteristic 'humpy' shape to the side, with a horizontal symmetry axis, consisting only of straight lines (check it here).

Now, in linear A we had a sign, labelled LinA *305, that has a fairly similar layout: Consising of straight lines only, having a horizontal axis of symmetry, with a fair degree of resemblance to the greek letter capital sigma ('summa sign'). (If you want, you can check its details here)

Now what if this identification is correct, and the sign LinA *305 is indeed the same as LinB *72? It is obvious that LinA *305 is primarily not an ideogram, it is too frequently used as a component of words. We can directly check it in the Linear A texts (thanks to John Younger for that). On the HT9 tablet,line b2, we see KA-*305, a term contrasted to the term SA-*315 in the first line, on the other side. Since we may observe a term KA-PA as a very common transaction term on the Haghia Triada tablets, it is more than tempting to assume a reading KA-PE for HT9b.2. As for the other occurrances of *305 as a phonetic sign, HT10a3-4 gives the reading PE-RU, HT27a.4-5 KU-PE (this may be connected to KU-PA, another word appearing on the tablets - e.g. on ZA11), HT146.3 RI-PE, Kh7a.4 SE-PE, PE2.2 RU-PI[]PE-MI, PHZb48 QA-PE[], ZA6a.1 PE-WA-NA, etc.

The same method can be applied to the Linear B sign *68 (RO2). It is a 'lyre-like' sign, with vertical symmetry axis, and has an inner cross that is already pretty much like the bare RO sign LinAB *02. I would propose an identity of the LinA *68 with the LinA *315 sign. The latter has similarity to LinAB *02 as well, though it is not always perfectly symmetric with a vertical axis. Again, LinA *315 is more of a phonetic sign than a logogram. Now, if we use this value to read Linear A texts, we may get some interesting results (the truth is, that the identification of these occurrences of LinA *315 with the phonetic value RO has already been proposed by John Younger). But let us not simply subsitute RO to these places but RO2 (which might have been *RYO). This way, we see the word SA-RO2 on the mentioned HT9 tablet. This reading does not seem imaginary, since we have a very common transaction term on the tablets: SA-RA2. Another nice thing is to observe, is that SA-RA2 never alternates with *SA-RA. This reminds us that the reading of RA2 was different from that of RA (it is often assumed to be RYA). But the term KU-RO can alternate with the term KU-RA (e.g. ZA20). In turn SA-RA2 seems to alternate with SA-RO2.