27 Oct 2011

Small quibbles about Proto-Berber orthography

Phoenix responded to a minor issue I raised about Proto-Berber orthography in Why I reconstruct *β and not *v. In defense of using a relatively arcane symbol  (taken from the IPA system) for a v-like sound that could instead be accommodated by a straight-forward symbol *v, he supplied the following reasons:
  • "In African linguistics v is commonly used as the symbol for the voiced fricative while β is used for the labial approximant."
  • "So I don't use v to transcribe Proto-Berber β, because it would suggest that it is the fricative counterpart to *b."
So from what I can see, his justification for the specialist symbol boils down to phonetics and tradition in the field. However I fail to find any justification here grounded in a clear methodology of some kind.

To the first argument, I suggest that basing an orthography on the phonetic level is inevitably cumbersome because it's then prone to constant revision as new discoveries about underlying phonetics come into view. A more stable and sensible orthography is based on the higher phonemic level instead, which focuses less on exact articulation of each sound in its context but instead displays for us *distinct* sounds of the language. For example, in English, the phoneme /p/ is pronounced differently in "spun" than it is in "pat". The /p/ in the former example is completely without a puff of breath (ie. [p] in IPA symbols) since it follows /s/ while in latter example, /p/ is indeed pronounced with a puff of breath by default (ie. [pʰ]). However on the higher phonemic level, we represent in both examples the single phoneme /p/ to eliminate extra irrelevancies that are ungermane to the focus at hand. It'd be likewise unnecessary to write out every word of a proto-language like Berber with only phonetic symbols rather than phonemic ones unless the topic was specifically about the exact articulation of each sound.

It's also a fact that there are exceedingly few if any languages that contain two distinct phonemes /β/ (bilabial fricative, pronounced by blowing through near-closed lips) and /v/ (labiodental fricative, pronounced with the lower lip touching one's upper teeth). It's pointless to obsess on minutia about the exact articulation of the sound if it can be reasonably ascertained that the sound was v-like. It then suffices to take advantage of an available letter from the Roman alphabet, *v, to aid readability both by specialists and by people in general. Things should be written with clarity for both specialists *and* the general public when possible lest it encourage ivory tower attitudes, the scourge of current academia.

To the second argument, tradition indeed is a seductress but it must be rejected when it no longer clarifies but obfuscates. Sometimes tradition is misguided. Sometimes tradition is outdated. Sometimes tradition is just plain wrong. In this case, I feel that this tradition is wrong precisely because of the first argument, that orthographies should reflect the phonemic level not the phonetic and that by ignoring this rule, one has unnecessarily obfuscated rather than clarified.

Possible solutions

After reading Phoenix's explanation with deep interest, I pondered on how the system might be revised to be clearer and to follow a more consistent methodology in its design. By following the principle of phonemics over phonetics, and by reserving diacritics and special symbols for the rarer sounds of a language marked by special articulatory features, we can arrive at a more balanced and clearer phonology.

Breaking with empty Berberist traditions, emphatic sounds may be marked by the underdot, as in Proto-Semitic studies. Again, we all may quibble about the exact pronunciation of (or *q) but a revised symbol  has the definite advantage of visibly showing a shared feature of "emphatic" with the other emphatics which would likewise be indicated more consistently with the dot: *ḍ*ḍ (former *), *ġġ (former *qq), * and *. The missing emphatic counterpart of *b, represented in this new system as **ḅ, is now impossible to confuse with non-emphatic *v which lacks the underdot. We may finally eliminate unnecessary IPA symbols and replace them with more generally readable symbols from the standard Roman alphabet that we already use while simultaneously making explicit any shared features that the different sounds may have in the language, such as "emphaticness".

And finally, through this revised system, specialists may continue to debate on the exact articulation of *ġ and such, but it won't affect the symbol shared among the specialist community until the phoneme's emphatic nature or its existence is disproven.

UPDATE
(1 hour later)
Upon further thought (my mind never stops!!), enforcing a surface representation with unvoiced letters might be even more kosher and, again, this would be even more in line with what's done in Proto-Semitic linguistics. So alternatively, we could use the following symbols to clean things up: * (= *), **ḳ (= ), *ḳḳ (= *qq), * (= *) and *ṣṣ (= *).

26 comments:

  1. Good explanation and I'll definitely take it into consideration when I get around to rewriting Proto-Berber phonology.

    Fun fact: Some dialect of middle atlas Berber actually have for Proto-Berber *ḍ.

    I'm still not completely convinced on the argument in the use of *v though. You identify it as a v-like sound, but I think this is open to debate.

    To me, as a Dutch speaker, the Bilabial approximant rather sounds like a w-like sound, since it's relatively close to our Dutch w [ʋ] a labio-dental approximant.

    But we can't use w but I'd like to express in some way that it was probably an approximant just like *y and *w but we've run out of plausible Latin symbols to express an approximant.

    I think I'll have to agree that, if your first and foremost priority is to use up the Latin alphabet before you resort to something more exotic, v is your best option, especially considering that at one point in time, it was used to represent a (bi?)labial approximant in Latin.

    Another option, which is employed by quite a few Berber linguists still is to write it as *h. It is quite clear that this approximant lost its labial articulation in a lot of languages. In Tuareg the *v has become h almost universally, in Zénaga it has caused vowel length. In most other languages it has (partially) disappeared. In Aujila and Ghadamès it's a bilabial approximant though. But this correspondence has been much overlooked, because these languages are often forgotten about (despite their enormous importance to reconstruction of Proto-Berber).

    ReplyDelete
  2. Phoenix: "Fun fact: Some dialect of middle atlas Berber actually have ṭ for Proto-Berber *ḍ."

    Then this fact is sufficient to avoid using *ḍ in favour of *ṭ. This aligns well with Proto-Semitic studies as well as the ejectives in Ancient Egyptian (as shown by Loprieno).

    "To me, as a Dutch speaker, the Bilabial approximant rather sounds like a w-like sound, since it's relatively close to our Dutch w [ʋ] a labio-dental approximant."

    Hold on, I detect some issues being confused here. Given an approximant *w already in the protolanguage, the "v-like sound" is unlikely to be a second, very similar approximant [ʋ]. Rather, we expect a more distinguishable fricative sound like [v] or [β]. Many languages use [ʋ] only as an allophone of /w/, as is the case in Mandarin and many European languages. Surely it's most sensible that it's a fricative, no?

    And yet, even if we could prove beyond doubt that it was a bilabial fricative [β] (as found in Spanish or Greek) instead of a labiodental fricative [v] (as in English), *v remains the clearest choice of symbol despite.

    We must avoid dragging the matter of orthography into phonetic debates.

    And I'd bet the average English-speaker would mistake [ʋ] for /v/ more often than /w/. This surely has to do with the differing relationships between phonemes when speaking English or Dutch. Proto-Berber speakers would've had yet a third point of view based on the internal structures of their phonological system.

    "Another option, which is employed by quite a few Berber linguists still is to write it as *h."

    Yet as you've conclusively pointed out, Aujila and Ghadamès make that tradition unwarranted. (That is, h doesn't just become a v wholesale without a clear trigger. That's absurd. Surely it can only sensibly be vice versa.)

    So we're nonetheless left with a v-like sound which may be most clearly represented simply as *v in orthography while β may be reserved for more exact phonetic transcriptions *only* when the need for this higher level of detail arises.

    "I think I'll have to agree that, if your first and foremost priority is to use up the Latin alphabet before you resort to something more exotic, v is your best option [...]"

    It certainly *is* our priority considering that everyone in the Western world has learned in gradeschool their ABCs (ie. Roman alphabet). It is THIS alphabet that we start with before deviating into the more complex and obscure. Unjustified complexity is corruption.

    ReplyDelete
  3. Surely it's most sensible that it's a fricative, no?

    I don't think so. As you can see in my table of reconstruction, only the dental series has a 'fricative series' *s, *z, *ṣ besides it's stop series *t, *d, *ṭ.

    If *v was a bilabial fricative and not an approximant, you would expect two series in the Labial field as well:

    **p, *b, **ṗ and *f, *v, **ḟ (that last character is an f with a dot above, just in case fonts make that unclear).

    A lack of **ṗ and **ḟ is unsurprising due to typological rarity of such sounds, I would still expect *p, *b, *f, *v. I agree that it doesn't take much imagination to imagine the *p to *f shift, especially when surrounded by Arabic like Berber is.

    So that *v is a fricative, is still a possibility

    The thing that finally made me me settle on an approximant rather than a fricative are the reflexes of this sound in the Berber languages.

    *v is often lost, but is still sometimes found. In the vicinity of an *u in many languages *v becomes ww.

    In coda position it causes vowel lengthening of schwa to *i or *u

    In the position *avər it often becomes awr And in Tashelhiyt *Cvə yields Cwə.

    I think these shifts are more likely if you assume an underlying approximant than a fricative.

    Finally, there's the question whether languages exist with two such apparently phonetically close approximants as [w] and [β̞]. I'll fully admit that the answer to that is 'I don't know'.

    I would have to check in what way Lanfry describes the Ghadamès sound v (which he transcribes with ). He might call it a fricative, he might call it an approximant, or maybe he isn't clear on the subject. But that, then, would be one language that has both.

    I took a look at UPSID (http://web.phonetik.uni-frankfurt.de/upsid_find.html) and found exactly one language that might have the distribution that I'm suggesting for Proto-Berber. This is obviously problematic. (http://en.wikipedia.org/wiki/Lakkia_language)

    So maybe a fricative after all.

    --

    *v behaves differently from all other Proto-Berber consonants except for in that it cannot be lengthened to *vv, whenever *v is put in a position where it would be lengthened, it metathesizes away to avoid this situation.

    For example: Aor. *ănvəy 'to see' in the Imperfect should have a consonantal structure C1ăC2C2ăC3, but instead it has a consonantal structure C2ăC1C1ăC3: not *năvvăy but *vănnăy.

    There is one other consonant that does not occur in a lengthened form and that is . This, oddly, seems to suggest that *v is the fricative counterpart of the glottal .

    Proto-Berber avoids the **ʔʔ in different ways. Medial in the verbal system is eerily rare. but when you do find a word, it generally has the alternative Imperfect formation tt-ăC1C1ăC2C3.

    The (near) absence of a medial makes you wonder if it also underwent metathesis, where it was then expanded across the whole paradigm by analogy.

    The absence of a *vv is an issue, whether you reconstruct it as a fricative or as an approximant. I'm very curious if you can think of something to explain this odd gap.

    ReplyDelete
  4. Phoenix: "you would expect two series in the Labial field as well: **p, *b, **ṗ and *f, *v, **ḟ [...]"

    Symmetry isn't a rule. It's a guide. By making this argument, you abandon considerations from historical linguistics. Berber comes from Afro-Asiatic and is related to Egyptian and Semitic. Nothing in Proto-Afro-Asiatic studies warrants our speculation on an opposition of *ṗ and **ḟ in the first place.

    In fact, Egyptian's phonology appears to evade symmetry too. The marked counterparts to the voiceless series of *p, *t and *k were naturally *b, *d and *g. Yet it wasn't a strict "voiced series" or "ejective series" because while *b was fully voiced, *d and *g were voiceless and glottalized.

    The lack of symmetry is related to the tendency for labial ejectives to weaken faster than other ejectives in a language, a *real* principle that even helps to explain Proto-Indo-European's unsymmetrical lack of **b.

    Without *p in the system, all likelihood is that Pre-Proto-Berber *p became *f. This common change occurred also in Etruscan and Celtic. This leaves *b and *v/ to analyse.

    You say that "in coda position it causes vowel lengthening of schwa to *i or *u" but a voiced fricative can cause such lengthening.

    However considering the curious change of schwa to front *i rather than simple assimilation to rounded *u and your grammatically motivated impressions that *v is the fricative counterpart of the glottal , I start to wonder if something was in the way between the vowel to be lengthened and the quality of the sound itself. Something like a glottal stop?

    This would suggest a third theory, that isn't so much "v-like" as an ejective bilabial stop *ṗ, the one expected based on Afro-Asiatic comparisons. It may be reasoned then that in Pre-Proto-Berber coda positions was pronounced as preglottalized [ʔp] contrasting with coda-final *p [ʰp] (pre-aspirated) and *b [b] (voiced). The weakening of glottalization here would lead to laryngealization of the consonant and preceding vowel, explaining the lengthening and raising of that vowel. With the glottal component between the vowel and consonant, rounding assimilation would be resisted as schwa rose to high-front *i.

    Following this hypothesis to its logical conclusion, our expected **năvvăy, being reanalysed as **năṗṗay exposes a problem since a preglottalized *ṗ in the coda would lead absurdly to **[naʔp.pʔay] ("." marks syllable boundary), strongly motivating a change.

    Yet a fourth alternative I offer is that Pre-Proto-Berber *ṗ devolved to *ʔʷ, ie. a labialized glottal stop and glottalic variant of plain *w. Yet again, the glottal segment explains the lengthening and rise of schwa to *i through coda-final laryngealization. Expected **năvvăy, being reanalyzed as **năʔʷʔʷăy, would motivate a simplification to **năʔʔʷăy before becoming *ʔʷănnăy, paralleling the metathesis you likewise observe with the plain glottal stops.

    ReplyDelete
  5. Interesting hypotheses.

    I'm not too familiar with Proto-Afro-Asiatic reconstruction, and I feel we still have a long way ahead of us in reconstruction of Proto-Berber before we can even start seriously reconsidering its position in PAA.

    But since you've caught my interest I must ask, would good PAA cognates with *v work within the *v = *? hypothesis?

    *ulv 'heart' must correspond to Semitic *libb 'id.'; Are there Egyptian cognates to this?

    Sadly, very little PAA cognates that I know about are accurate when it comes to *v because Militarev (http://starling.rinet.ru/cgi-bin/query.cgi?basename=\data\semham\brbet&root=config&morpho=0) made a serious mess out of it by interpreting it as *h and kind of winging it along the way. (For some illustration see: http://lughat.blogspot.com/2010/07/unreliability-of-afroasiatic.html) Needless to say, we pretty much need to rethink any half-hearted eyeballed sound-correspondence that Militarev thinks to have found.

    Let me just list a couple of very basic Proto-Berber words with *v maybe they'll ring a bell for you in Egyptian/Semitic. My somewhat limited knowledge of semitic gives me absolutely nothing.

    *a/a?vu 'smoke'
    *ev?/a? 'night'
    *a??v 'milk'
    *a?r?v 'to write' (probably 'to scratch', but means 'to write' in a surprising amount of Berber languages)
    *avarn 'flour' < Lat. farina (?)
    *ev?n 'tent'
    *tiv?li 'sheep'
    *tivatt-en 'sheep' (pl.)
    *tav?nt 'stone'
    *tavurt 'door' < Lat. porta (?) This word actually has a lengthened vv in Aujila tavvurt and also in Ghadamès taßßurt. This is the only word that has that.
    *av?/aya(w) 'cousin, grand child' (it actually means both in several languages).
    *uva/a?r 'lion' The initial vowel of this noun is extremely uncertain.
    *ta??uvt 'wool'

    ReplyDelete
  6. "*ulv 'heart' must correspond to Semitic *libb 'id.'; Are there Egyptian cognates to this?"

    Yes! Its Egyptian counterpart is transcribed ỉb (which I presume is for actual *ib). It lost former *l-.

    "Let me just list a couple of very basic Proto-Berber words with *v maybe they'll ring a bell for you in Egyptian/Semitic."

    Nothing pops out at me so far, but then Berber added prefixes that even Egyptian lacks so I'll think on it and see if I find anything. I'm deeply curious myself about the developments of AA *ṗ.

    ReplyDelete
  7. Oh wait, we might be able to peck off "lion" which might be regarded as a loan; compare Egyptian rw (for *rīwa ?) and Central Chadic *ruw- 'hyena'). It looks like metathesis unless there's a better etymology available.

    ReplyDelete
  8. Ugh something went horrible wrong with my previous post.

    Here's the words again:

    *ă/aʔvu 'smoke'
    *evə/aṭ 'night'
    *aḳəv 'milk'
    *ăʔrəv 'to write' (probably 'to scratch', but means 'to write' in a surprising amount of Berber languages)
    *avărn 'flour' < Lat. farina (?)
    *evən 'tent'
    *tivəli 'sheep'
    *tivatt-en 'sheep' (pl.)
    *tavənt 'stone'
    *tavurt 'door' < Lat. porta (?) This word actually has a lengthened vv in Aujila tavvurt and also in Ghadamès taßßurt. This is the only word that has that.
    *avə/ăya(w) 'cousin, grand child' (it actually means both in several languages).
    *uvă/aʔr 'lion' The initial vowel of this noun is extremely uncertain.
    *taʔṭuvt 'wool'

    ReplyDelete
  9. Here's another brainstorm. Let's search for a sound that contains a labial quality (to account for the Ghadames and Aujila reflexes), a glottal quality (to explain it's behaviour parallel to plain glottal stops), no voice and frication (to explain the curious Latin correspondences with /f/ despite already having *f in the traditional system). And given *aʔvu we'll have to concede that a value of /ʔʷ/ for our problematic *v doesn't quite cut it.

    This might lead us to /hʷ/, which can sound much like /f/ in Latin. It's labialized, it's glottal, it's voiceless. But in order for us to accept that, it seems like we need to shift pre-existing *f to something else like the otherwise missing *p.

    Now *ahʷărn would correspond to Latin farina, *tahʷurt to Latin porta. The latter correspondence is a dilemma unless our new *p (former *f) tended towards a lack of aspiration.

    ReplyDelete
  10. By the way, *avə/ăya(w) 'cousin, grand child' reminds me of Proto-Semitic *ˀaḫū 'brother', particularly if we could pin the value of *v to /hʷ/ as per my last suggestion.

    ReplyDelete
  11. "By the way, *avə/ăya(w) 'cousin, grand child' reminds me of Proto-Semitic *ˀaḫū 'brother', particularly if we could pin the value of *v to /hʷ/ as per my last suggestion"

    That'd be a promising correspondence. But if you want to uphold the correspondence with *libb ~ *ulv we're going to have to set up some conditioning of a PAA soundchange to accomodate PB *v corresponding to PS *b and *ḫ.

    I should add that not all Latin loanwords with have f corresponding to *v.

    ifilku 'fern' < Lat. filix 'id.'
    tafurka 'catapult, branch' < Lat. furca 'fork'
    afalku 'beareded vulture; falcon' < Lat. falco 'falcon'

    I'd venture to say that Lat. f more commonly corresponds to *f than to *v. The two possibly Latin loanwords that I have suggested up there are far more uncertain than the ones where Lat. f corresponds to *f.

    Lat. p also often corresponds to *f afullus 'chicken' < Vulgar Lat. pullus 'chicken' < Lat. pullus 'little animal'
    ifirəs 'pear' < Lat. pirus 'pear tree' or pirum 'pear'.

    There are clearly multiple strata of Latin loanwords though. as you can see *afullus has a perfectly classical latin -us ending. but *urtu < hortus 'garden', *muru < murus 'wall' lack it like vulgar Latin.

    Maybe we can find a sort of stratification

    Stage 1: Berber loans Lat. f with *v, and Lat. p with *f (which was probably still *p at the time, we have latin transcriptions of Berber tribes names where p corresponds to *f. Latin ending -us is still intact.
    Stage 2: Berber loans Lat. f with *f and p with *f. Latin ending -us has become *-u.

    I don't think there is enough Latin loanwords to prove such a hypothesis though.

    Interesting fact: *avṣalim 'onion' is a Loanword from Hebrew bəṣālīm 'id.'
    That loan is probably from Punic not Hebrew, and must have been much earlier than the Latin loan.

    Why loan Hebrew b with PB *v?
    Did PB *b represent a different sound at the time?
    Or was it loaned from a context where Hebrew had lenition like: li-ḇəṣālīm 'to/for the onion'?

    Needless to say, the last thing has not been said on this subject yet.

    ReplyDelete
  12. Berber will not make a monkey of me! I shall be victorious! ;o) My adaptation to new data will not cease.

    "But if you want to uphold the correspondence with *libb ~ *ulv we're going to have to set up some conditioning of a PAA soundchange to accomodate PB *v corresponding to PS *b and *ḫ."

    Rössler:1952 related the Berber form to these cognates and presumed *ulh < *luh < *lub < *lubbu. Personally I wonder if *u- here isn't the masculine marker somehow, leaving the biradical *-lv to contend with. I've been voraciously adding Berber roots to my offline database and have been noticing a lack of word-final *b. Could this explain the change somehow?

    Also *ˀaḫū 'brother' may be from earlier *ˀaḫwu so that *ḫw would trivially become *hʷ (or alternatively we might write *w̥, if you will) in the earlier forms of Proto-Berber. Is this sufficient?

    "There are clearly multiple strata of Latin loanwords though."

    Yes, and so we might postulate Proto-Berber 1 having *p and voiceless *w̥ (= [hʷ]) at first. At this stage, the closest approximation of Latin /f/ available is *w̥.

    In Proto-Berber 2, *p would shift to *f, leaving the odd gap while *w̥ would be drawn to *v (= [β]). Now the closest approximation to both Latin /pʰ/ and /f/ is indeed *f (= [ɸ], former *p).

    "Why loan Hebrew b with PB *v?"

    Are there phonotactic constraints in Proto-Berber against not just word-final voiced plosives (*b, *d and *g) but coda-final?

    ReplyDelete
  13. Ooops, I might have jumped the gun on phonotactic constraints at the end, given *aˀfud 'knee' and *avigg 'colt'. Sorry. I retract. But I still have no word ending in *-b curiously and I do wonder if phonotactics are the cause of this Hebrew/Berber correspondence somehow.

    ReplyDelete
  14. Hahaha, correcting more of my mistakes above. I meant to type "biliteral *-lv" above, not "biradical". Sigh. My brain's not cooperating today.

    ReplyDelete
  15. Good catch on the lack of final b!

    Kossmann (1999:127) cites 4 words with a final *b, but almost all these words are problematic. Thefore Kossmann concludes:
    "Il n'est donc pas impossible que le proto-berbère n'avait pas de distinction *H (our *v) ~ b à la fin du mot".

    It does occur in coda position though. For example *abrid 'chemin'. *v often becomes b in coda position, but they're well distinguished in Ghadamès and Zénaga.

    ReplyDelete
  16. Thanks, that's the power of having your own searchable databases. And I also noted *abrid which seemingly shows coda-final *b... or does it?

    Could *abrid also be parsed as */a.brid/ rather than */ab.rid/? If so, I see no other instance of coda-final *b so far.

    Coming back to the AA "heart" correspondance then, *b lenites to *v in *ulv. Is it from Pre-Proto-Berber *wa-lib? Who can say?

    ReplyDelete
  17. "Could *abrid also be parsed as */a.brid/ rather than */ab.rid/? If so, I see no other instance of coda-final *b so far."

    It's hard to come up with a criterium to determine this.

    Best way to determine syllabification is by looking at where schwa is inserted.. but since this differs per language, that doesn't help much.

    Best way, but hardly conclusive argument on the syllabification is that geminates universally spread over two syllable *-tt is */-t.t-/. Many Berber languages will simply lose the length in an initial position, or insert a schwa in front of a geminate.

    "
    Coming back to the AA "heart" correspondance then, *b lenites to *v in *ulv. Is it from Pre-Proto-Berber *wa-lib? Who can say?"

    with short *a and short *i, yeah probably something like that. Sadly we don't really understand yet why some words have initial *u.

    With *urə/ăḳ 'gold' it's quite obvious that it's a derivation from *WRḲ 'to be yellow' thus *u represents a vocalized *W somehow

    Pre-Proto-Berber: *wariḳ ? Why don't these *u type words have a *a- masculine prefix?

    In case of *ulv we don't have a root **WLV 'to be a heart' obviously. We have a Iwelemmeden Touareg reflex: əwəl pl. iwăllăn which clearly shows a w. Does this mean the w was there still in Proto-Berber? It'd be the logical conclusion if it weren't for the fact that we can't get a consistent correspondence of initial *u in other languages and initial *əw in Iwelemmeden Touareg.

    The issue on *w and its vocalisation is far from settled. verbs of the type *ăkkər take a caustive *ssukər. All of a sudden *u shows up and the geminate is lost. Prasse reconstructs *ăwkər, but that doesn't quite account for the causative which, in that reconstruction would be expected to be *ssăwkər, why does *ăwk yield ăkk and *ssăwk, suk?

    All questions that I hope to find answers to one day. As always, looking at more data tends to help a lot here, so I'll collect more data!

    ReplyDelete
  18. "It's hard to come up with a criterium to determine this."

    I refuse to be stumped now. ;o) We can learn from *avṣalim that there's little reasonable way to parse it but */av.sˤa.lim/ since **/a.vsˤa.lim/ is absurd according to the general theories of sonority hierarchy and syllabicity. This shows that morpheme boundary doesn't coincide necessarily with syllable boundary.

    So it's at least within the realm of speculation that *abrid is to be parsed syllabically as */a.brid/ rather than */ab.rid/. We could attempt a testable rule in Proto-Berber regarding syllable parsing such that, say, the only legal syllable structure is C₁(C₂)V(C₃) with an avoidance of coda whenever possible. With *abrid, we can avoid a coda in the first syllable but not in *avṣalim where we coincidentally see the softening of expected *b.

    "Best way, but hardly conclusive argument on the syllabification is that geminates universally spread over two syllable *-tt is */-t.t-/."

    This would work with my on-the-fly syllable rule so far.

    "Pre-Proto-Berber: *wariḳ ? Why don't these *u type words have a *a- masculine prefix?"

    First, I've seen that word before: Egyptian wȝḏ and Semitic *√wrq 'to be green'. Loprieno reconstructs Egyptian */ˈwaːʀiɟ/ 'green'.

    Second, if we attempt a reconstruction of Pre-Proto-Berber *u-lib 'heart' becoming *u-ləb then *uləv (lenition of coda-final *b), then a Pre-Berber form *wuriḳ- 'something yellow/green' becomes likewise *wərəḳ then *urəḳ 'gold'. This follows regular rules where AA/Pre-Berber *u and *i merge to mid-central while *a remains low. It's a thought I had at least.

    Third, did Pre-Berber have word classes like in other African languages to the south? In which case, a tentative *u-lib would simply belong to a special noun class with *u-marking. Often body part terms are lumped into special classes if I remember my Swahili grammar.

    Finally, if these sound changes are sensible, it's curious that the vowel reduction is reminiscent of (but not identical to) the reduction in Pre-Egyptian as if by possible areal influence *after* Egyptian and Berber were already distinct languages.

    "[...] *ăkkər take a caustive *ssukər."

    Okay I'm stumped for now. ;o) Back to the drawing board.

    ReplyDelete
  19. Oh drats. In keeping with my rules, I guess that should be Pre-Berber *wu-lib or *wi-lib in order to get to *uləv. I'm flying on the seat of my pants on this one, I know.

    ReplyDelete
  20. That *ăkkər/*ssukər alternation is driving me nuts. I have to solve it tonight so that I can sleep, hehe. Is it perhaps possible that what we have here is a case of *a-Ckər/*s-Cəkər? If the missing consonant were *v, we'd have *a-vkər/*s-vəkər. It just seems to me like some missing consonant is causing gemination. That's the only thing I can dream up at the moment. There, g'night everyone! Zzzz...

    ReplyDelete
  21. Yes! Go get your sleep.

    As for class marking: I'm sure people have tried to pin the prefix system to that so they wouldn't have to think how it derived from PAA, but it's not going to help much. It's not like Swahili where you can actually find semantic classes with certain prefixes.

    I guess if you want to play devils advocate something could be said for *i being employed for quite a few body parts. But you know that the word for both hand and leg don't play along.

    A lot of prefixes only combine with certain root vowels. To get a preliminary indication of how it works, you should have a look at my Master Thesis that I uploaded on academia.edu.

    There are four prefixes: *a-, *e-, *i- and *u. *a seems to originally come from an indicative pronoun. I expect *e- to be an allophone of *a- (*e- is almost exclusively found in *eC(C)eC and *eC(C)əC forms. *aC(C)əC exists, but aC(C)eC doesn't. Smells like allophony doesn't it?

    I'll go into a bit more detail on my blog soon.

    The *u- prefix seems to be from an original *w-

    The *i- prefix is 1. not that common and 2. There is no indication that it would come from an initial *y.

    As for the gemination and degemination of caustives:

    I doubt anyone doubts that there's gemination due to a consonant causing it, even if it were just because we like triradical roots.

    The question is which consonant was that? *w seems like a promising candidate, because it 'explains' the *u in the perfect. But really it gives more trouble than it solves. What we should also keep in mind is that these verbs have vowel alternation. Let's assume there was a *C, its ablaut pattern would have looked like this:
    Aor. ăCkǝr Pf. ǝCkăr
    Aor. *ǝss-ăCkǝr Perf. *ăss-ǝCkăr

    We would need a form that would change the sequences *săC and *sǝC into *u, while the sequences *ăC and ǝC to cause gemination with the next consonant.

    I can't think of a sound exotic enough to behave like that. Haha. So maybe we should look in the direction of older vocalism patterns that were lost due to metathesis. Maybe normal verbs where CVCVC instead of VCCVC. But then we're still not really out of the woods yet. Suffice to say, this is a big (and certainly unsolved) problem in Proto-Berber reconstruction.

    Just to add: Verbal nouns of the C1C1C2 verbs like *kkər (or a much more universal verb *kkəs 'to take away') generally have the structure *uCəC too. But considering the irregularity of verbal noun formations, and the uncertainty in reconstruction of *u in nouns, it is hardly helpful.

    ReplyDelete
  22. Considering that you yourself knew no clear answers, I thought it rational to brainstorm. All I know is the extra *u- in 'heart' is present in neither Egyptian nor Semitic and appears to be a Proto-Berber innovation. So I couldn't help but reflect on a sensible possibility that neighbouring Bantu languages have perhaps helped shape (Pre-)(Proto-)Berber at any point during the several millenia we're examining. Can we honestly object to this if we have so much trouble explaining a tiny prefix in one word?

    "The *u- prefix seems to be from an original *w-"

    Or more specifically *wV-, implying that one way or another there is an unexplored condition in Pre-Proto-Berber where *wV- can be reduced to *u-. We need to know what that condition is rather than remarking idly about it. I've attempted ideas. Your turn.

    "The *i- prefix is 1. not that common and 2. There is no indication that it would come from an initial *y."

    I notice Kossmann reconstructs the PBerb 3ps prefix as *yǝ- which develops later into i-. There's no question that it relates to Semitic *y-. Since you say that the prefix *i- on the other hand is uncommon, it makes me wonder how you can be so certain that it cannot be from earlier *y- despite this sparse data you admit to.

    "To get a preliminary indication of how it works, you should have a look at my Master Thesis that I uploaded on academia.edu."

    You mean Nouns of the CVC and CC type in Berber? It's excellent! There's a lot of information here to absorb and I'm still reflecting on it.

    I'm curious though about the extra glottal stops in places that seem to be disconnected from Egyptian and Semitic cognates that I'm aware of. You cite the forms *iɣasseˀ 'bone', *elseˀ 'tongue' and *aḍawdeˀ 'finger' which, if I'm not mistaken, relate to Egyptian qs 'bone', ns 'tongue' and ḏbˁ 'finger'. Here are three Egyptian words which have no trace of these extra glottal stops that you posit for Berber. Can you explain why?

    "I can't think of a sound exotic enough to behave like that. Haha. So maybe we should look in the direction of older vocalism patterns that were lost due to metathesis."

    Yes, it's definitely a fun puzzle but I don't think gemination requires an exotic sound. Compensatory lengthening suffices oftentimes. On the other hand, I suppose I'm a little uneasy with claims of grammatical metathesis when so many Nostraticists and long-rangers have abused this idea to death.

    Note that Kossmann concedes that the sound in *ăčəˀ 'eat' may be the result of a Pre-Proto-Berber cluster *ty, implying that the imperfective *tăttăˀ, if I've understood correctly, was geminated by way of a weak consonant.

    So I'm still intrigued. I'm allured too by the symmetry behind the notion that gemination can be caused by weak consonants while at the same time they're barred from gemination of their own.

    ReplyDelete
  23. "I notice Kossmann reconstructs the PBerb 3ps prefix as *yǝ- which develops later into i-. There's no question that it relates to Semitic *y-. Since you say that the prefix *i- on the other hand is uncommon, it makes me wonder how you can be so certain that it cannot be from earlier *y- despite this sparse data you admit to."

    Proto-Berber 3ps masculine is *y-, not *yǝ-, schwa insertion sometimes happens epenthetically, but in general the first vowel after *y- is part of the stem.

    It doesn't just develop into *i- universally. It is dependent on the phonetic environment, in most languages it works like this:
    *yǝ-CC > yǝ-CC
    *yǝ-CV > i-CV

    Results are different especially in Tashelhiyt, where vocalization depends in a set of rules based on sonority (a r is sooner to vocalize than a k even if the k is in a traditional place to get a schwa).

    This is a significantly different situation than nominal prefix *i- which can appear both in -CC and -CV environment. Therefore a *yV reconstruction is probably not fruitful.

    Showing the *u isn’t from *wV is considerably more difficult. We can show that *wă- yields u-/wǝ- in Northern Berber languages in the same distribution as *y- prefix in the verb by looking at the état d’annexion of nouns with initial *a. But we cannot show that the other short vowel behaved in the same way in that context.

    You know what, this is becoming an incredibly long discussion. I think we’ve zeroed in on a subject which deserves its own blog article, or maybe even several articles.

    “Where does the *u, and *i come from as prefixes?”
    “How does the root initial *w behave in Proto-Berber?”

    [Comment cut in two pieces, blogspot doesn't allow huge replies apparently]

    ReplyDelete
  24. "you cite the forms *iɣasseˀ 'bone', *elseˀ 'tongue' and *aḍawdeˀ 'finger' which, if I'm not mistaken, relate to Egyptian qs 'bone', ns 'tongue' and ḏbˁ 'finger'. Here are three Egyptian words which have no trace of these extra glottal stops that you posit for Berber. Can you explain why?"
    That is great news! As I point out in my thesis: "Having examined the words presented above [those here you cite here]. It can be concluded that all words with a final glottal stop in my sample all present problems that make it difficult to reconstruct the Proto-Berber form. It is striking that the three words discussed at length here all have a glottal stop that seems to be lost in the plural of Zénaga.”
    In other words, whatever the reliability of the glottal stop in the singular, it oddly disappears in the plural.
    I’m not sure why. I’ve entertained the thought that the specific clusters these words created called for an epenthetic i in Zénaga which is indistinguishable from , except that the glottal stop will disappear the instant that it isn’t necessary anymore, as seen in the plural.
    This works quite well, and it’s not hard to imagine that Zénaga would need an epenthetic vowel to facilitate a final cluster tʸš. Something similar can be imagined in a final cluster ʔss for ‘bone’.
    But the solution is ad hoc. There’s not an inherent problem with word final clusters in Zénaga. Pointing these two clusters out as the ‘environment’ to cause final glottal stop will probably yield you these two words, and no others at all.
    Does that mean it isn’t true? No. Does it make it impossible to prove? Yes.
    I’d say that Proto-Afro-Asiatic helps us out here. By showing that there’s no glottal stop, specifically in these words, might be extra reason to believe that the glottal stop in Zénaga is indeed a Zénaga-internal development.
    Are there other words where the glottal stop is more established where it does correspond perfectly to a glottal stop in Egyptian?

    For ‘finger’ I simply have no idea. Zénaga has a difficult final cluster γḏ which it may have wanted to break up by adding a final i. It’s just that no other language has this cluster (Zénaga γ usually comes from *g). Most languages don’t point to a *w at all, Ghadamès points to *kk, which cannot be a result of *ww or something similar. Aujila points to *qq and Zénaga points to *g or *gg. It’s an incredibly problematic word.

    ReplyDelete
  25. "Are there other words where the glottal stop is more established where it does correspond perfectly to a glottal stop in Egyptian?"

    A clear cognate with Berber is to be found in Egyptian *ˀanāka (> Sahidic anok) and Semitic *ˀanāku. (Loprieno has cited *janák with ahistorical *j but the reed leaf symbol, normally representing y, here represents the word-initial glottal stop since *y- was not possible in the language, as in Minoan.)

    The matter of AA comparison with Egyptian is complicated by the fact that former *r has also become glottal stop in Egyptian (eg. Pre-Egyptian *wārigʸ- becomes *wāˀaḏ 'green').

    "For ‘finger’ I simply have no idea."

    The Egyptian correspondence, with that of "heart", has me beginning to wonder if a more accurate form in Berber might not be more along the lines of *aḍavḍ.

    ReplyDelete
  26. I'll fess up: I'm having troubles with what the Pre-Egyptian form should be for "green", *wārigʸ- or *wārikʸ-. Perhaps the latter if the sound is originally voiceless in AA. Anyways, you get the jist.

    ReplyDelete