20 Feb 2010

Aegean phonotactics against word-initial /j/

On the Minoan Language blog, Andras Zeke counters my entry against a prefix *i- in Minoan with a new idea that the morpheme in question was a separate deictic instead. Clever, however only I-QA-*118 (HT44) ~ QA-*118 (KH 10) and I-DA-MA-TE (AR Zf1) ~ DA-MA-TE (KY Za 2) are available as evidence for this vocalic utterance, only significant if we assume that the two items of each pair have identical meaning. The pair YA-SA-SA-RA-ME (TL Za 1) ~ A-SA-SA-RA-ME (PR Za 1) only suggests a consonantal onset but this then is poor evidence for a purported morpheme that's independent of the word it precedes. Instead, if we assume that the pairs with I-/∅- are proper, the disappearing initial *i- is likely an unaccented vowel prone to syncope. This leaves just YA-SA-SA-RA-ME ~ A-SA-SA-RA-ME to explain, which isn't likely to be due to some disappearing morpheme, there one minute and gone the next.

In Prefixes in Minoan, I alluded to the Proto-Aegean phonotactic constraint against word-initial *y- which is evident not only in Minoan but in Etruscan as well. Of the 1504 words and names logged in my Etruscan database, not a single one shows an underlying word-initial /j-/. There's no question to me then that Etruscan phonotactics simply barred the glide from word-initial positions altogether. The apparently random alternation of YA-SA-SA-RA-ME ~ A-SA-SA-RA-ME in Minoan can be most parsimoniously explained by the same constraint. Given this feature in the language, symbols for YA and A would naturally be interchangeable in word-initial positions, as well as the pairs YE/E, YU/U, etc.

Aegean languages aren't the only ones to do this and I'm starting to see an interesting pattern around the Eastern Mediterranean. Egyptian too wrote (for consonant /j/) in places where /j/ no longer existed. So while the name of the god Amon was spelled out as ỉ-m-n in the vowelless script, the name was pronounced *ʔAmúna in the second millenium BCE (hence Coptic amoun). In other words, like I suggest for Linear A, Egyptian script wrote phantom glides in word-initial position while having developed a phonotactic constraint against word-initial /j/. More on this in Egyptian writing systems and grammar [pdf] by Shawn Knight.

Only recently did I have a brainwave about Greek and its development from Proto-Indo-European (PIE). Many IEists historically have noted that PIE *y- becomes either h- (PIE *yēkʷr̥ 'liver' > ἧπαρ hepar) or z- (PIE *yes- 'to seethe' > ζέω zéō). I never thought of this innovation in the context of this Minoan constraint rule before but I should have thought of this earlier in hindsight. Perhaps this is food for thought?

Then there are the hints in Anatolian languages of a possible similar constraint that I have to look into, at least where *ye- is concerned.[1][2][3] All in all, there appears to be a geographical pattern of linguistic areal influence here with Minoan at the center.

[1] Münchener Studien zur Sprachwissenschaft, Issue 62 (2006) (see link): "However, the lack of any sure reflexes of */ie-/ in either Palaic or Lydian precludes the possibility of certain CAnat. status for this conditioned yōd-loss. (MELCHERT AHP 1994a: 75, KIMBALL HHP 1999: 361-2)."
[2] Melchert, Anatolian historical phonology (1994), p.75 (see link): "Hittite, Luvian, and Lycian give evidence for loss of initial */y/ before */e/. There are no certain examples for */ye-/ in Palaic and Lydian, so attribution of this change to PA must remain tentative."
[3] Woodard, The Ancient Languages of Asia Minor‎ (2008), p.42 (see link) concerning Palaic: "The absence of examples of initial /y/ is surely accidental, but the lack of initial /r/ is systematic, as elsewhere in the ancient Anatolian languages." (identical to Melchert's statement in 1994:206).


  1. This is an interesting point of view - as I noted before.

    I also realized that I did not cite all the possible examples for the initial I-/J-, so I amended that with the addition of I-TI-TI-KU-NI / TI-TI-KU and the all-too-well-known JA-DI-KI-TE-TE- / A-DI-KI-TE-TE- pairs. Yet I must admit the scarcity of Minoan finds make it difficult to draw a decisive conclusion. I only chose to interpret I-/J- as a separate particle, as I saw some pattern in its alternation on the Libation Formulae, and its occurrances on tables. But given the low number of finds, no such theory can give 'significant' results (in a statistical sense). [So I cannot prove nor disprove a theory on this ground].

    It's really off-topic, but I wonder what you can make out from the I-QE particles suggested in the same post. Does it seem linguistically nonsense for you or a possible thing to see a relationship with Etruscan Iχ ? I tried hard to read I-KA, but that would really violate the natural affinity of signs with Linear A (it is very-very hard to see KA in the place where QE 'gives itself').

  2. Bayndor: "I also realized that I did not cite all the possible examples for the initial I-/J-, so I amended that with the addition of I-TI-TI-KU-NI / TI-TI-KU"

    The whole point of minimal pairs is that they have only one contrasting feature. Your example above has two (I- and -NI). It's therefore rejectable from a strict methodological standpoint.

    "So I cannot prove nor disprove a theory on this ground"

    The onus is yours and Occam's Razor sides with the least amount of assumption. You assume that I- is a morpheme without having proved it. Given the facts so far, the default solution remains that I-/J- is not a morpheme at all until proof is offered.

    "Does it seem linguistically nonsense for you or a possible thing to see a relationship with Etruscan Iχ ?"

    In our ignorance of what language this represents and what it signifies, of course anything is possible. I can't find a rational objection to the proposal, if that's what you're inquiring.

  3. Perhaps I should also add this warning about the Phaistos Disc: Times Online: Phaistos Disc declared as fake by scholar. I can't say the thought hasn't crossed my mind many a time myself.

  4. I know all too well about Occam's Razor, being a scientist (albeit not in linguistics). However, this is a two-edged blade: The low frequency of identifyable Minoan words with regular (or irregular) alternations plagues just about any theory: I am afraid that also includes yours. Since you proposed a regular phonological development of initial glides (JA -> A), some spatial or temporal pattern in the A <-> JA alternation would be handy to support this theory: But the truth is, the same words were written with J- and without J- by people in the same geographical area and the same age. To show a real development (and not just some wild variation), it would be especially neat if at least a single example of a word evolving from that of a JA- initial could be shown (such initials are quite common in Linear A - I counted more than 20 different words with JA- initials - roughly the same number as the words with I- initials give).

    Anyway, since it is not my fashion to dispute a theory without proposing a new one supported by at least some evidence, I went further on my path to search for the origins of this mysterious 'I'-deictic. I think it is possible to theorise the existence of such a particle purely on the basis of Etruscan words (so we do not even need the evidence from other Aegean languages). Take the examples of the Etruscan demonstrative pronouns IKA and ITA. They also exist in forms that lack the initial I-. Perhaps it's just because I grew up on Latin texts, but it does remind me the case of the Latin demonstrative pronoun hic/haec/hoc. In that case, a pronoun (*hi) was merged with a separate deictic (*ke) to give hi+ke >hic. But the merger was incomplete, so there were regular declination forms lacking the *ke particle. It is very tempting to see a somewhat similar development in the case of Etruscan pronouns: the merger with an initial *i- deictic. And I also know where this *i deictic might have come from. Perhaps it's just madness, but take a look at the 3rd person singular personal pronouns: it is IN (for objects) or AN (for persons). We also know that Etruscan words usually lacked the accusative marker, save the case of pronouns, where the suffix *-n was used. But the 3rd person personal pronouns look like as if 'frozen' into an accusative case. If it were true, the original nominative forms would have been *i for inanimate nouns and *a for animate nouns. The word *i = "that" would make a perfect and natural deictic, that could later have been merged onto the demonstratives KA and TA. (It would be interesting to explore whether *a has seen any use as a deictic for persons).

    As for the Phaistos disc, I do not think it is a forgery: and just because of one thing. You are right about it, that if it were unique, then its rejection would be a natural conclusion. But the characters of the Phaistos disc are not unique: there exists a single other document with similar signs: the famous Arkalochori axe. The signs of the Arkalochori axe are not Minoan Hieroglyphics. Though not entirely identical to those of the Phaistos disc, they are somewhere in-between the Disc and Linear A. For example, the 'plumed head' (=I) is seen in positions that are apparently word-initial; the 'cat head' (=MA) sign is fully drwan out, but in the frontal direction more usual in Linear A; and finally the 'Column' signs (=NA) have a dotted shaft, more similar to Linear A, but their headpieces otherwise look like the Phaistos disc 'Column' signs (and - in this regard - they are dissimilar to the Heroglyphic 'Column' signs). The DA signs are drawn as full twigs, like those on the Disc, etc. I do not think that too many scholars would declare the Arkalochori axe as a forgery, too. And if it is not a forgery, then it shows that the signs of the Phaistos disc fall well within the 'normal' variations expected for the Minoan writing systems, especially Linear A.

  5. Occam's Razor is never a two-edged sword. It may only seem that way if you're on the wrong side of it and are looking for an easy way out.

    "Since you proposed a regular phonological development of initial glides (JA -> A)"

    This isn't what I proposed. I proposed that the JA/A alternation is purely orthographic due to a lack of phonemic contrast between [#ja] and [#a], as is attested also in neighbouring Egyptian. The loss of *y- is necessarily dated to a time BEFORE Proto-Aegean proper since this feature exists in both Minoan and Etruscan.

    "The word *i = "that" would make a perfect and natural deictic, that could later have been merged onto the demonstratives KA and TA."

    Strange, I've posted exactly about this pre-Etruscan *i- deictic and its relationship to animacy, ergativity, and PIE *i- before online somewhere (Yahoogroups like Cybalist perhaps?); you must have read it.

    Even so, there's no semantic sense in adding demonstratives in the places you require to make your idea work. Why "THE Asasarame" one minute and simply "Asasarame" the next? Courting with randomness isn't a scientific virtue.

    "As for the Phaistos disc, I do not think it is a forgery: and just because of one thing."

    Yes, fair enough. Only cracking the Minoan language from the other extent artifacts will truly reveal this answer. I'd hardly throw the object out of my museum (if I was blessed enough to have one) simply because of a lurking possibility.

  6. Thanks for the link about the Phaistos Disk. Unfortunately, such fakery has not been unheard of.

  7. I was just reading an article by Alwin Kloekhorst (http://www.kloekhorst.nl/KloekhorstLydianDatSg.pdf) suggesting that the Lydian λ is originally from "i̯". It's worth noting that a palatal glide is conspicuously absent from Lydian. Seems like an areal feature with the neighbouring Aegean languages.

  8. Along with Egyptian, we might also include Elamite to the list of languages treating word-initial /j/ as a non-phoneme.

  9. I found another quote concerning Egyptian. On page 38 of Allen's The ancient Egyptian language: An historical study (2012):"These data have given rise to three opposing interpretations: that j was originally /y/ and became /ˀ/ in most words; that it was originally /ˀ/ and became /y/ in a few cases; or that it was bivalent, representing both /ˀ/ and became /y/ (reflected in the alternative transcription ı͗)." Here, the author overlooks a fourth possibility, that the palatal glide [j] was simply disallowed in word-initial position altogether, serving to neatly explain the first person pronoun ı͗nk as /ˀanāka/ without an unetymological glide (ie. there simply is no evidence for this pronoun ever having initial /j/ when compared with Proto-Semitic *ˀanāku).