Phoenix's blog

Oriental Berber

The collaborative blog focused on the Berber languages of eastern north Africa is now live!

Look forward to many translations and articles focusing on these fascinating languages.

http://orientalberber.wordpress.com/

Hope to see you there!

Multiple reflexes of final PB *əv

There are not many cases of Proto-Berber final *əv in Aujila (only two). These two have two different reflexes.

ĕrní [v.]
imp. sg. ĕrní pl.m. ĕrniyât pl.f. ĕrnîmet; pf. 1sg. ernîḫ 3sg.m. yernî; pf2. 1sg. ernîḫa 3sg.m. yernâya; aor. 1sg. a-rnîḫ 3sg.m. a-yérni; impf. 1sg. rennîḫ 3sg.m. irénni 3pl.m. rennân ‘to increase, add; to bid’

cf. Ghd. ărnəβ ‘to add’; Kb. ərnu ‘to add’; Zng. aṛīh ‘to increase’

The loss of *n in Zénaga is striking, but the other two cognates give enough basis to reconstruct:

PB *ărnəv

arév [v.]
imp. sg. arév (rarely: úrev) pf. 1sg. uréfḫ 3sg.m. yurév; 1sg. urífḫa, urívḫa 3sg.m. yurîva; aor. 1sg. a-urévḫ 3sg.m. a-yúrev; impf. 1sg. tārévḫ, tāréfḫ 3sg.m. itârev ptc. târęven ‘to write’

Berber, cf. Ghd. órəβ ‘to write’; Kb. aru ‘to write’; Zng. īrih, iʔrih ‘to dictate’

The *v in Zénaga reflects as vowel length, but vowel length and a glottal stop seem to be incompatible in this language resulting in two forms, one with metathesised *v and one with the glottal stop in place (Ghadamès o also points to an initial glottal stop).

PB *ăʔrəv

So we have two cases of final *v with two different reflexes. They're different verb classes in most Berber languages, but that shouldn’t really affect the final consonant reflex.

I have no idea how to explain what exactly is going on here. But I thought it was interesting to point out.

Perhaps ĕrní was loaned from a nearby Berber Language with a reflex of *v, Siwi for example, although I haven’t found this word in Siwi. The word doesn’t look like that likely a candidate to borrow anyway.

CC* verbs in Aujila

As my work on the Aujila language continues, I try to figure out some of the syntax and morphology.

Part of every Berber language’s verbal morphology is the way it handles the CC* verbs, or CCa/i verbs. These are verbs with two root consonants and a final vowel that alternates in different contexts.

Historically, these verbs were triradical verb whose final consonant was a glottal stop. The resulting paradigm and subsequent analogies quite radically different paradigms in the different language, but one theme can be distilled. Almost always the variation of vowels varies between a/i/ø depending on the person and tense of the verb.

Ayer Tuareg (Niger) is an exception to this, which merged all CC* verbs with the CCu (<*CCh?) verbs. Figuig is somewhat special too as it has variation with a/i/ø/u.

Kossmann (2001) discusses this variation and the reconstruction of these Verbs in Proto-Berber. I think it is a worthwhile addition to discuss how Aujila handles these verbs, as it is quite exotic in some respects.

There are three ‘groups’ of vocalism that can be distinguished. The 1sg. and 2sg. always group together with the same vowel. The 3sg.m., 3sg.f. and 1pl. always group together with the same vowel. Finally the 2pl.m., 2pl.f., 3pl.m. and 3pl.f. all take the same vocalism. These groups don’t seem to be any different in Aujila, although it is sometimes hard to tell, since the available data especially lacks 2pl. forms (A common problem in any language where the only access to the language is textual, 2pl. forms just don’t occur very often in stories).

Due to these groupings it is enough to cite just the 1sg. 3sg.m. and 3pl.m.:

• aor. 1sg. cca-x
• aor. 3sg.m. yə-cc 
• aor. 3pl.m. unknown. So far, I have not encountered a single aor. 2pl.m., 2pl.f., 3pl.m. or 3pl.f.
• perf. 1sg. cci-x
• perf. 3sg.m. yə-cca
• perf. 3pl.n. əcca-n
• impf. 1sg. cəc̄i-x
• impf. 3sg.m. i-cəc̄a
• impf. 3pl.m cəc̄a-n

Not all of these forms are special. The perfect behaves similar to most other Berber languages (Tashelhiyt, Ghadamès, Ahaggar Tuareg, MA Berber; But not Riffian and Figuig).

The Imperfect behaving the same as the perfect is interesting though. In MA Berber, for example, the final vowel is always a, regardless of the ending. In Riffian there is usually no vowel at all.

The final, most striking feature is the 1sg. and 2sg. of the aorist having a vowel a. This is unlike any other Berber language. The origin of this development is quite obscure as well.

The /x/ in Aujila does have a colouring effect on /ə/ which sometimes results in Paradisi transcribing what should phonologically be an /ə/ as <a>, <ä> or <å> (e.g. pf. 3sg.m. yáḫzer [PT:VI,VII], yäḫzér [PT:III], yeḫzér 'to look' for /y-əxzər/). I wondered if it might be that the regular ending was -əx and for 2sg. t-...-ət (as with most languages) and that the -əx was coloured to -ax creating an imbalance. In the perfect and imperfect the 1sg. and 2sg. are grouped together, while in the aorist. due to colouring the two forms would have two different vowels. This would then be spread by analogy to the 2sg.

But, this explanation would only work if /ə/ shifted to /a/ phonemically, and I doubt that this is the case. There are many examples where <a> that should be a /ə/ still behaves as a /ə/. For example pf. 1sg. saḫḫ /səġ-x/; pf2. ssíḫḫa /síġ-x-a/ ‘to buy’. Here we see the typical raising of /ə/ to /i/ when the -a suffix pf the pf2. is attached.

So, this analogy-explanation is not particularly attractive. This leaves us with a vowel /a/ that I do not quite understand.

I’m curious to hear how exactly Siwa Berber handles these types of verbs.

---

I decided to try a new font for my blog, it should now display in Gentium 12pt. The default font my blog theme uses is not a Unicode font, giving quite an ugly look when using Linguistic symbols. Gentium is a lovely Oldstyle serif font, with the support of a lot of funny characters us linguists like to use.

If you don’t have it already, you can get the font here. While Gentium Plus is better at stacking diacritics, which is sometimes important, the default Line spacng is much broader which I don’t like, it also seems like the hinting is not as good.

I have also justified the text, instead of flush-left. The standard justification engines of operating systems are quite terrible, so I’m not sure if I’ll continue to do this. But the text field seems just broad enough to make it work and not make it look too horrible.

Let me know what you think

---

Maarten Kossmann (2001) L'origine du vocalisme en zénaga de Mauritanie. Frankfurter Afrikanistische Blätter, 13, pp. 83-95

I mark the resultative form with pf2. for convenience, as it is a derivation of the perfect. See my previous blog post on the formation of this form.

Ghadamès Riddles 5

Yá mən ḥayátək εaf—ažărtil iṭkár s-imaẓawăn, wăl illi dó-s xaf yón n-ărγa.

‘A mat filled with fresh dates, there is only one yellow date.’

• ažărtil ‘mat’ [Lameen:] ažărtil is widely borrowed into North African Arabic; it appears in Korandje as zartalli.
• iṭkár 3sg.m.pf. ‘to be filled'
• s-imaẓawăn ‘with fresh dates’
• wăl negative particle, which according to Lanfry’s description is not supposed to combine with the negative perfect. Quite odd.
• illi 3sg.m.neg.pf. ‘to be, exist’
• dó-s in-it
• xaf ‘except’
• yón (m.) n-ărγa litt. ‘one of yellow date’

<ōū> = /əw/

Sometimes, language presents evidence so clearly, you are kind of surprised that it works.

While working out the Aujila phonology, I ran into severe problems with Paradisi's transcription, which is incredibly precise on a phonetic level, but for phonological analysis is mostly a lot of unecessary clutter.

One of the sequences I had trouble dealing with was what Paradisi transcribes as <ōū> (which in the Italian tradition is used for a short diphthong, while <ōu> is used for a long diphthong, I swear I'm not making this up.), to make my system 'work'I needed it to be read as /əw/, while this is of course possible, but you have to bridge a gap from phonetics to phonology to actuall show it.

The proof I found presents itself in the verbal conjugation of Aujila. Aujila shares a tense with Siwa Berber which Lameen Souag calls 'relevance', I prefer to call it 'resultative', it is a derivation of the perfect, which expresses an action finished in the past, of which the result is evident or relevant in the present. To me the function seems to be almost identical to the Greek Perfect.

Despite Aujila and Siwa being quite different phonetically, somehow one of the two languages succesfully transfered this innovative tense to the other language. Citing from Lameen Souag's PhD thesis (pg. 389): "A feature of the system [...] is the marker -a; this is places at the end of the verbal word, following any subject or indirect object agreement markers or direct object pronominal suffixes. If the form to which it was suffixed would otherwise have ended in əC, the ə changes into i" (emphasis is my own)

The crazy thing is, the above description is just as valid for Siwa as it is for Aujila, including the i-insertion.

Because of this, the <ōū> = /əw/ correspondence can now be established clearly.

perfect. 3sg.m. yeddér /yəddə́r/'he lived'

resultative 3sg.m. yeddîra /yeddír-a/ 'he's lived (and the result of this is obvious/this is relevant to the present)'

---

perfect 3sg.f. terṓū /tərə́w/ 'she gave birth'

resultative 3sg.f. tirîwa /təríw-a/ 'she's given birth (and the result of this is obvious/this is relevant to the present)'

And there we have it <ōū> = /əw/.

Let's jump into the time machine!

A while ago, one of my professors at Leiden University, Michiel de Vaan, was asked to help out with an episode of a new television series on Belgian television about language. He was asked to write a small conversation in Dutch as it was spoken 500, 1000 and 1500 years ago, and to explain what kind of development the language has gone through during that time.

The resulting episode is quite nice. I hope my non-Dutch readership will still be able to pick up some of the stuff that is being said. Because it is absolutely worth checking out, also, because it isn't particularly time consuming.

Het Nederlands van eeuwen geleden

If there's anything to complain about is that de Vaan uses Proto-Germanic for Dutch of 1500 years ago, I'm not very familiar on this stuff, as Germanic isn't my specialty, but there seems to be enough evidence that around 500 there must have at least bean a little bit of differentation, but who cares, it's cool.

As a result of this episode, the Dutch artist Spinvis decided to write a small song forming a small dialogue between old Dutch and modern Dutch.

Ick ne mag thi kussjan

Ghadamès Riddles 4

wăǵǵid inăqq əddrari-nnăs, Al as săllon ălžiran-ənnăs

'A man (who) kills his children, the neighbours (are) hear(ing) him.'

• wăǵǵid 'a man'
• inăqq 3sg.m.impf. 'to kill' An interesting, quite irregular verb, the aorist and perfect are ănn and ənno, with assimilation of nγ to nn, the γ show's up again in the imperfect, but this time geminated, and thus automatically becomes qq.
• əddrari-nnăs 'his children'
  
• Al the particle of the progressive aspect which combines with the imperfect. It attracts clitics of the following verb. Progressive aspect does not translate too well here, Lanfry also suggests that simultaneity of two actions can be expressed. This might work better.
• as the attracted 3sg. indirect object clitic.
• săllón 3pl.m.impf. 'to hear'
• ălžiran-ənnăs 'his neighbours' < Ar. žīrān 'neighbour'

Ghadamès Riddles 3

Yá mǝn ḥayátǝk εáf—tălta ak tǝxǝddǝm, Asíd tăβdǝd εáf-éγăf-ǝnnăs.

'A woman that does not work until she will stand on her head'

• Most of this formula is the same as last riddle.
• tăβdǝd 3sg.f.aor. 'to stand upright'
• εáf- 'on', a common Berber preposition, with an irregular (but quite common) γ > ε shift (in IPA that's [ʁ] > [ʕ]).
• éγăf-ǝnnăs 'her head', with an amazing etymological repetition. εáf- 'on', is derived from this word 'head'.

Ghadamès Riddles 2

Yá mǝn ḥayátǝk εáf—tălta ak tǝxǝddǝm, Asíd ttar ḍarăn-ǝnnăs

'A woman who does not work, until she opens her legs'

• Yá mǝn ḥayátǝk εáf— Standard opening for riddles. I cannot make any sense of it. Is it Arabic?
• tălta 'woman', used to signify that the word this riddle is looking for, is a feminine word, different from last riddle where wăǵǵid 'man' was looking for a masculine noun. Quite interesting that they are able to personify grammatical gender so well.
• ak 'negative particle'. Surprisingly not the particle wăl or something similar that we find in most other Berber languages is used (wǝl, ur, ul etc.), the cognate of that negative particle is only used for the negative imperative.
  
• tǝxǝddǝm 3sg.f.neg.impf. 'to work'
• Asíd 'until'
• ttar 3sg.f.impf. 'to open'
• ḍarăn-ǝnnăs 'her legs'

Notes on Aujila Phonology

Now that I've discussed all of the Aujila texts. It's a good time to start considering the Phonology of Aujila. In this post I'll try to give a general outline on what my preliminary conclusions are.

There are two general points of discussion, first is the consonant system second is the vowel system. The consonant system is not very difficult to analyse. The vowel system is considerably more difficult and Paradisi employs a staggering amount of vowel signs, I count more than 20. These vowel signs have been employed to mark very minute phonetic details. Paradisi has not attempted to write the vowels phonemically, and thus it is quite difficult to figure out the phonology of the vowels.

Consonant System

In general I think we can formulate the consonant sysem as follows:

• Labial: f, pp;  b, bb; v, vv; m, mm; w, ww
• Dental: t, tt; d, dd; ṭ,ṭṭ; (ḍ); s, ss; z, zz; (ṣ, ṣṣ); ẓ, ẓẓ; n, nn
• Palatal: š,šš; ž, žž; y, yy
• Velar: k, kk; g, gg; x, xx
• Uvular: q, qq; ġ
• Pharyngeal: (ḥ) ; (ʕ)
• Glottal: (h)

The phonemes in brackets only appear in Arabic loanwords. q is mostly found in Arabic loanwords, some people think the Aujila q corresponds to Proto-Berber *g, but I think there's very little unproblematic evidence for this. qq is the long counterpart to ġ In native Berber words. 

The most exciting part of the consonant system, from a historical linguistic perspective is the perfect retention of the Proto-Berber *v (*h or *β if you prefer).

Vowel System

For some time, I had the impression that Aujila retained the Arabic short a, but all of the cases where this seems to happen, it's within an emphatic environment, and we must conclude that it is simply a colored ǝ.

For the Berber vocabulary I propose a 4-vowel system. a, i, u, ǝ. The Arabic loanwords add several vowels. Most importantly e and o. There may be a short ŭ, I'l discuss the problematic vowel later on.

Depending on the environment, Paradisi wrote the four vowels with radically different spellings. Below, I give an overview of the different writings of the vowels and in what environment these variant appear. This overview will not be exhaustive, but it will give a general idea.

• Stressed a:
  • â in non-initial syllable.
  • á in initial syllable.
• Unstressed a:
  • ā in non-initial syllable.
  • a in initial syllable.
• Stressed i:
  • î in non-initial syllable.
  • í in initial syllable.
• Unstressed i:
  • ī in non-initial syllable.
  • i in initial syllable.
• Stressed u:
  • û in non-initial syllable.
  • ú in initial syllable.
• Unstressed u:
  • ū in non-initial syllable.
  • u in initial syllable.
  • Somtimes: ů

This looks quite systematic. But the fact is that in initial syllable, especially the unstresed variant, we often find both the non-initial and initial variant. Sometimes find the initial variant in a place where we would expect the non-initial variant too.

• Stressed ǝ: Usually: é, sometimes: ẹ́.
  • In emphatic environments: Usually á, also ǻ, ắ, ä́. Therefore, Stressed ǝ in emphatic environments often overlaps with Stressed a in transcription.
  • Stressed ǝww: ṓū
• Unstressed ǝ: e, ẹ.
  • ĕ in initial position. (and initial syllable?)
  • Before a nasal: ę. Sometimes I find ę in other positions, and e, ẹ are found in front of nasals too.
  • In emphatic environments: ä, å, ă, e, ẹ, and rarely a.
  • Unstressed ǝww: ōū
  • In front of y: i
  • In front of w: u

e and o in Arabic Loanwords seem to only occur stressed, and are thus written ê and ô.

There are several cases where Arabic short u corresponds to either u or ú in Aujila. While this transcription seems to suggest a short pronunciation, one has to wonder whether this is a phonetic distinction, since it is not in the Berber words.

This vowel is typically found in verbs from C1C2C2 verbs in Arabic. For example laffa, iluffa 'to wrap up' becomes illúff in Aujila.

It may simply be that Paradisi was hearing shorter vowels, because he expected them due to his familiarity with Arabic. Another option is that ú is an alternate spelling for û in front of geminates. We have a nice indication for this in the Berber vocabulary as well. 'knee' has two variations: afûd and afúdd, sadly the origin of this geminate in this word is completely unclear.

This 'in front of geminate' explanation has on exception: ṣṣundûq 'box, chest', which I expect is a loanword from Classical Arabic/MSA

So there you have it. A fairly unsurprising system for a Berber language, except for that one exotic retention of Proto-Berber *v.