6 Feb 2009

Novilara Stele remains a mystery

A few days ago, I received a mail concerning the North Picene language and the Novilara Stele. The language appears to have been present in NE Italy around the first millenium BCE. The stele in question reads thus:

mimnis . edut . gaarestades
rotnem . uvlin . partenus
polem . isairon . tet
sut . trat . nesi . kruvs
tenag . trut . ipiem . rotnem
lutuis . θalu . isperion . vul
tes . rotem . teu . aiten . tasur
soter . merion . kalatne
nis . vilatos . paten . arn
uis . balestenag . ands . et
ut . iakut . treten . teletau
nem . polem . tisu . sotris . eus
I was asked whether this language looks related to Etruscan or Rhaetic languages and in my honest opinion it's not. While it appears to have an Indo-European-like look, it continues to defy translation. We might note some grammatical patterns in it. The ending -ten marks a number of words here (aiten, paten, treten) and is similar to a second person plural ending found in PIE languages. The words teu and eus remind me vaguely of the Indo-European pronouns *tu "you" and *n̥s "us". Some words appear to be marked by an accusative singular ending -m as we typically find in Indo-European languages (rotem, polem) and others appear to be marked with the tell-tale Indo-European nominative ending *-s (gaarestades, lutuis, ands, sotris).


  1. Looks like the ending may be -em, not just -m. Every word ending in -m also ends in -em. I think I'll have more comments soon.

  2. So I've been doing a little research on this. There doesn't seem to be much information out there. What's worse, no one can seem to agree on a transcription. Some sources transcribe g as c. The word edut is also transcribes as erut. There are two different s sounds. In one case, kru(v)s, both letters are found. There's also ambiguity over whether one of these s letters should be transcribed as m instead. It's unclear if partenus is one word or two, parten us. There's also the question of whether iakut is actually lakut. There's an uncertainty with merion~merpon. Another uncertainty is gaarestades~gaaristades. Finally, which word breaks are indicated in a line, it's unknown if they're indicated when a word is broken up between two lines. So there's lots of questions, and few answers.

    With that caveat there are some things I've noticed. The vowel o appears only before dentals (l, n, t, and s). The other vowels appear before dentals so there's no complementary distribution. It's not clear why that should be. The vowels a, i, and u appear to be about equally common, but e is much more common. I hypothesize (i.e. wildly guess) that *o > e except before dentals.

    Most letters appear word initially either 2 or 3 times. In contrast to this, t occurs 10 times. This suggests a t-prefix of some sort. Furthermore, initial t never precedes either u or o. The lack of to- could be explained by the general rarity of o, but u is actually more common that i, so the absence of tu- is curious. I do note that s (the sibilant) occurs word initially only before o. Similarly, s (the shibilant) occurs word initially only before u. Thus the two s sounds could be allophones of t in word initial position.

    I think we can get rid of the shibilant as a phoneme. It occurs only 4 times. Twice it is word initial before u. Once it appears in the ambiguous reading kru(v)s (it can be a sibilant or a shibilant). And once it appears in the word rotnes~rotnem. Since rotnem and rotem are both attested I'm inclined to think the correct reading is rotnem.

    Word finally t, s, n, and m are all common and it's likely that they are suffixes (or in the case of m it's really -em that is the suffix, and possibly -ut instead of just -t). Another obvious suffix is -ten. This sequence is very common. -ag has been suggested as a suffix as well. It only occurs twice, but both times it's after the sequence (-)ten-. The first time in the word tenag and the second time in the word balestenag. It's possible that this could be a compound of the words bales and tenag. It's also possible that tenag is a suffix (or two). If that's the case then I'm inclined to think that the word tenag is actually a suffix on the preceding word kru(v)s. This points to an agglutinative structure. Balestenag may be broken up into bale-s-ten-ag.

    Tomorrow I'll see what I can do with word order.

  3. Some phonotactical observations.
    • Boustrophedon is out of the question since rotnem is found on both lines 2 and 5.
    • The rarity of k g is striking, only 3 counts each (vs. 34 of t) and all of them adjacent to a or r.
    • I agree with the conclusion that -ten is a suffix. Possibly -rion as well.
    • Also, the m : n ratio is 1 to 6 in onset but 8 to 7 in codas, so there is probably a suffixal -(e)m too.
    • All but one of the complex codas (nds vs js rn) end in -s, it could be a suffix too.
    tenag and balestenag must be related words.
    θalu stands out as the single aspirate; they could simply be rare (voiced stops are rare too), and since t is the most common stop, θ is exactly the one aspirate we should expect to find in that case. But is there any suitable proper name this could be?

  4. Two more things, Mordrigar.

    As is shown on Omniglot regarding Italic scripts, the only form of "u" used in North Picene is equivalent to the South Picene character transcribed as "uu". Whether or not this is important, I'm not sure...

    The other thing to point out is the last word of line 5, "rotnem/š/ši". The stele is chipped where the last character(s) would be, so the ending is not entirely certain. Even so, the sequence "rotne-" has already appeared at the onset of line 2.

  5. Has Dusan Polansky written you about the stele?

  6. You wrote: "A few days ago, I received a mail concerning the North Picene language and the Novilara Stele" so I wondered if he had written you as he had done to me a while back:

    Dear Ms. Gill, or anybody else reading this email,

    please find attached my text on Novilara stele. I am addressing you > in consequence of finding a nice article about Novilara stele on > your web.

    Text in the attachement is pure amateurism, PDF is not well formated.
    In other words the form is very poor, but the content is highly > explosive.

    It may be wrong in details, but I am sure of being right in the core.

    Feel free to forward these files to anyone who could be interested > or possibly publish any part or whole text by any means.

    Best Regards
    Dusan Polansky

    I can't evaluate it, but would gladly forward it to you, if you like.

  7. So let's see if we can figure out anything for word order. Most languages on the planet are either SOV or SVO.I'm pretty sure that's true of this particular linguistic area as well. Since we're reduced to guessing, at this point, the best guess we can make is that the word order is either SOV or SVO. Let's try to break things down.

    mimnis . edut

    We start by assuming that this is SV. So -s is a subject suffix and -ut is a verbal suffix.

    gaarestades rotnem . uvlin

    The -s suffix appears again signaling the start of a new sentence. If this sentence has an object and a verb then we can't yet determine which of -em and -n is the verb and which is the object.

    partenus polem . isairon

    We're in the same boat here. Word order seems consistent though.

    tet sut . trat . nesi

    This sequence is interesting. The first word, tet (assuming this who section is a new sentence) appears to be the subject. Probably it's uninflected. If sut is the verb then apparently roots could consist of a single phoneme (this can happen in Kartvelian too, IIRC). I don't know what to make of trat.nesi.

    kruvs tenag . trut . ipiem . rotnem

    My guess is that kruvstenag is one word since -ten seems to be a suffix and -tenag is also apparently a suffix elsewhere. This would indicate an agglutinating language.This sentence also allows us to resolve the earlier issue of whether -em is a nominal or verbal ending. kruvstenag is the subject and trut is the verb so ipiem rotnem must be the object (there might be two objects, or perhaps ipiem is an adjective meaning that adjectives agree in case with their head noun and word order is adjective-noun).
    We thus deduce that the -n suffix is a verbal suffix.

    lutuis . θalu . isperion . vul

    Here the unknowns are θalu and vul. θalu might be an object and perhaps vul is an adverb or some particle.

    tes . rotem . teu . aiten . tasur
    soter . merion . kalatne

    This sequence is quite confusing. tes, rotem, and aiten all look like nouns and merion is apparently a verb. But what are we to make of teu, tasur, soter, and kalatne?

    nis . vilatos . paten . arn

    The first three appear to be nouns or some combination of nouns and adjectives. arn is apparently a verb.

    uis . balestenag . ands . et
    ut . iakut . treten . teletau
    nem . polem . tisu . sotris . eus

    The first three seem to form a noun phrase. et ut is possibly etut, a verb, which is followed by another verb. treten is a noun.It's also possible that teletau nem is one word, in which case it would be an object. tisu, like the earlier θalu, might be a noun of some kind (perhaps an indirect object).

    And to finish things off, stris.eus has the subject suffix -s, occurs sentence finally, and throws a monkey wrench into the works possibly invalidating everything I've just said.

  8. For the past week or so, I've been delving deeper into this stele, and have found a great deal more uncertainties than I expected.

    First, I tried to find as many relatively good-quality pictures of the stele as possible. I was disappointed to only find four legible pictures (http://img10.imageshack.us/img10/3459/novilara.jpg , http://www.cez-okno.net/files/clanok-subory-2010/Novilara_stele_front.jpg , and http://paleoglot.blogspot.com/2009/02/novilara-stele-remains-mystery.html), one only barely so (http://www.maravot.com/novilara_stele.jpg). Why are there not more pictures? Is flash-photographing the stele in its current state at the Museo Preistorico Pigorini not allowed? Or is interest in North Picene just that limited?

    Next, I inspected the inscription letter for letter, making a transcription, first using Paint (which took too long), and then Word, with the downloaded fonts "Cardo", "Damase" and "Aegean" (which was moderately faster, but not much). This got me a good 5 lines in; then, out of impatience, I just used a normal font, with {curly brackets} to denote variants.

    I then compared my transcriptions to others that I found online - primarily the one at Wikipedia (http://en.wikipedia.org/wiki/North_Picene_language) and Paleoglot (http://paleoglot.blogspot.com/2009/02/novilara-stele-remains-mystery.html). I used these as a basis for matching characters to letters.

    The letter transliterated here as “u” corresponds to the South Picene “ú” sign, but since it is the only “u” sign in the inscription, I have opted to omit the accent.
    San (the letter shaped like “M”) should technically be transliterated as /ś/ (s with an acute), it but is so prevalent in the inscription that I have opted for an accentless form.
    The letter transliterated as “š” differs from san by having a shorter left leg.


    The text is written right-to-left, and words are separated by hyphens (which, in some cases, are small enough to resemble interpuncts).

    LINE 2
    The last letter of rotne{m/š} has a long final stroke, i.e. at top left of the letter. It is possible that it is a damaged or misspelled “š”, but this seems unlikely, as the final stroke is clearly visible in 3 out of the 4 pictures. The most likely reading is rotnem.
    The presence of a dot in the “o” of rotne{m/š}, and possibly in polem and isairon (LINE 3), is almost certainly unintended, as it is absent elsewhere.

    LINE 4
    In {s/š}ut A very faint vertical line appears just below the rightmost stroke. The more likely reading is šut.
    The dash between trat and neši is very small.
    Some transliterations include an “i” in kruv{i}{s/š}. This seems unlikely: it would have to be illegibly crushed against the side of the last letter.
    Most of the last letter of kruv{i}{s/š} has been lost. An “s” or “š” is barely visible, but the telltale stroke is too unclear to give a solid identification.

    LINE 5
    The “a” in ten{a?}c is unusually angular, and the horizontal bar is unclear. Nonetheless, its identity as an “a” is very probable.
    The left side of the last letter of rotne{s/š/m} is damaged, leaving the phonetic value unclear.

    LINE 7
    The dash between tes and rotem is very small.

    LINE 10
    The dash between ands and et is very small.

    LINE 11
    The left edge of the first letter of {š/m/s}ut runs together with the second letter.
    The bottom of the first letter of {i/l}akut has been damaged, obscuring the telltale differentiating stroke.

    LINE 12
    The first letter of {n/w?}em is severely damaged, and practically illegible.
    Due to a deep chip in the middle of po{?}lem, a letter may be lost.

    The letter transliterated as “c” is highly unlikely to have the same phonetic identity as “k”, considering the minimal pairs caarestadeskalatne(nis) and balestenac – {i/l}akut. It may represent a voiced velar stop, as was the case for “C” in Early and Classical Latin.

    While there are only three definite uses of “š” (neši, tašur, and tišu), they occur in similar phonetic environments, i.e. preceding /u/ and /i/. If all eight possible cases are considered, six of them are succeeded by a front vowel:

    i | #e
    e | t
    e | (#)r
    u | (#)p
    i | a
    t(#) | u (probably š)
    v | (#)t
    e | (#)l
    i | #θ
    i | p
    e | #r
    r(#) | o
    i | #v
    o | #p
    i | #b
    d | #e
    u# | o
    u | #

    (e | #u) (probably “m”)
    t(#) | u
    e | i
    v | (#)t
    e | (#)l
    a | u
    i | u

    This tendency for “š” to fall before front vowels can be summarized in a rule:

    1) /s/ -> /š/ {_V(high)}
    The resultant “š” phoneme is likely realized with some variety of palatalization.

    There are two seeming exceptions to the rule: kruv{s/š}/ tenac and rotne{s/š/m}/ lutuis. The former is analogous in form to balestenag. Since the /s/ remains unchanged before a /t/ in balestenag, it is unlikely that it would be realized differently before the /t/ in kruv{s/š}/ tenac. Using the same reasoning, it is also likely to be a single word, kruvstenac.
    As for rotne{s/š/m}/ lutuis, it is certainly possible that /l/ causes the same change as a front vowel, but as we will see, it is irrelevant whether it does or not. Furthermore, the analogies of rotnem and rotem speak against identifying the final letter as “š” (unless different cases are at work here).

    Does this rule also apply across word boundaries? If so, then we are left having to explain why nis-vilatos fails to undergo the change. If not, we should accept that words are broken between two lines, based on tet/šut and possibly et/{š/m/s}ut.
    Going on the assumption that /s/ -> /š/ {_V(high)} does not apply across word boundaries, the likelihood that “š” is the uncertain letter of rotne{s/š/m}/ lutuis decreases even further.

    Having accepted that words can be broken between two likes, word boundaries at line ends are no longer certain. For what it is worth, the average word length is between 5 and 6 letters.

    mimnis erut caarestades
    rotnem uvlin parten us(-)
    polem isairon tet-
    šut trat neši kruvs-
    tenac trut ipiem rotne{s/m}
    lutuis θalu isperion vul(-)
    tes rotem tev aiten tašur
    soter merpon kalatne(-)
    nis vilatos paten arn(-)
    uis balestenac ands et-
    {š/m}ut {i/l}akut treten teletau(-)
    {n/w?}em po{?}lem tišu sotris eus

  11. Seadog Driftwood: "Why are there not more pictures? Is flash-photographing the stele in its current state at the Museo Preistorico Pigorini not allowed? Or is interest in North Picene just that limited?"

    I gauge that it's caused by a few idiotic things: antiquated copyright laws, ivory-tower elitism and human laziness, for starters.

    I believe that academics who are privileged enough to acquire PhDs owe back to the community and should be sharing their knowledge with the world, not hoarding it. I dislike secrets and dishonesty. If this is supposed to be the Information Age, then why aren't academics sharing more information? Perhaps they're afraid of intellectual competition but I'm all for it.

    "The letter transliterated as “š” differs from san by having a shorter left leg."

    No, you misinterpreted. The letter with the shorter leg is definitely "m" in order to distinguish it from san. This style is attested even in Etruscan records so you're paddling up the wrong creek. You may not be aware but san was used in Northern Etruscan to write plain /s/ while sigma was used for the esh-sound; this was the reverse of the southern preference.

  12. No, you misinterpreted. The letter with the shorter leg is definitely "m" in order to distinguish it from san.

    This objection confuses me. Surely the first letter of mimnis is neither the same as the last letter, nor the same as the penultimate letter of what I transliterated as neši and tišu! The letter transliterated as "m" has 4 diagonal strokes (like a modern "w" attached to the top-left of a vertical line); "s" has the more or less the shape as a modern capital M, with the leftmost stroke as long as the rightmost; and "š" is like "m", but with only 3 diagonal strokes.

    You may not be aware but san was used in Northern Etruscan to write plain /s/ while sigma was used for the esh-sound; this was the reverse of the southern preference.

    Fascinating! This explains why sigma does not appear anywhere on the stele, and only san is used.

  13. Forgot to clarify: I'm considering san to be /s/, and this odd sign that resembles either a san with a short left leg or an em with one stroke too few.

  14. Ah, I see the detail you're seeing now. You have an eagle's eye.

    However, for the sake of devil's advocacy, how would you debate against treating the different number of strokes in these "one-legged" letters as merely scribal variation for a single letter mu?

    Afterall, Egyptians wrote n (ie. the wavy water symbol) occasionally with five peaks, sometimes seven, sometimes more or less. The Modern Greek alphabet also shows variation of sigma depending on whether it's at the end of a word or not. Can we be sure that these considerations don't apply here?

  15. Someone noted that the sign R which occurs in the text twice, and which is transcribed as a d, looks like an anachronism. Neither western Greek nor Etruscan alphabets had such a letter. The western Greeks did have a P with a short downward stroke, but not a full R. It resembles the R in the Oscan alphabet, which however was somewhat later. Moreover, it's in Oscan that this letter stands for the sound d. So it seems curious, did the Oscan R descend from North Picene R? Or rather, is the Novilara stele a forgery? See the discussion here: http://www.antiqui.it/piceni/lingua.htm