7 Nov 2011

Changes in Pre-Egyptian vocalism

Lately I've been reflecting on what Loprieno says about the early Egyptian vowel system on page 55 of Ancient Egyptian: A linguistic introduction (1995):
"In our discussion of phonology (section 3.4.3), we saw that one of the major features of Egyptian in its early stages was the presence of a strong expiratory stress, which eventually caused a reduction to /ø/ of short vowels in open syllables in posttonic position, with the resulting change from the Dreisilbengesetz to the Zweisilbengesetz (**saḏimat > *saḏmat 'she who hears')."
While Loprieno speaks of reduction to zero, I've long been thinking more along the lines of a Pre-Egyptian system of *a*i and *u being reduced to *schwa* wholesale in all unstressed positions. To begin with, long vowels were only to be found in stressed positions in Pre-Egyptian, at least if the comparison with Proto-Semitic is trustworthy, and this length contrast in stressed positions clearly remained in Egyptian, as still evidenced by Coptic. I therefore choose to write all of these reduced, unstressed monophthongs of Pre-Egyptian as *a (to be implicitly understood as [ə]). Furthermore diphthongs *Vy and *Vw (*V = any vowel) then become *i [əj] and *u [əw] respectively. This has worked very well for me for a while now. The result is an Egyptian vowel system that still looks on the surface much like Proto-Semitic with long vowels restricted to stressed syllables and unstressed positions having only short *a*i and *u. Yet since the system has been notably altered, we find a curious incongruence nonetheless between the vowels of Proto-Semitic and those of Egyptian.

We can also avoid a lot of the wildcard symbols Loprieno and others occasionally use in the unstressed syllables this way since my theory makes this pointless: Only *a can exist in these positions unless accompanied by a written semivowel y or w in which case the appropriate short high vowel is selected. It appears that the matter of whatever the original vocalism may be is an issue for Pre-Egyptian reconstruction, not Egyptian proper. Loprieno's */'ri:ʕuw/ (> */'ri:ʕə/) 'sun' becomes my *rīˁa.

There are further reasons why I'm dwelling on this, but I've divided it up into subsequent posts.

49 comments:

  1. Interesting idea! As I see it, there are two important arguments in favor of non-stressed i and u: First, Akkadian transcriptions are usually interpreted as containing evidence for original nonstressed vowels. Second, the frequency of final -w being written depends on the preceding vowel.

    You idea would imply that unstressed vowels were "reactivated" when the accent shifted, as in the status pronominalis, feminine and plural forms. In many cases, these alternations (Coptic hor "face", hraf "his face") are used to determine unstressed vowels.

    But your idea seems nonetheless very interesting and worth following. Do you know Peust's monograph on Egyptian phonology? Might be an interesting read for you.

    ReplyDelete
  2. Thanks. Questions:

    1) Is it not possible that these vowels are merely the product of a scribe's perceptions of what was pronounced within the constraints of a cuneiform writing system which inevitably warped and reduced any foreign language to Akkadian's four cardinal vowels (a, e, i, and u)?

    2) Can you provide specific examples showing "the frequency of final -w being written depends on the preceding vowel." Which artifacts are you referring to? Are you referring to the Amarna numerals?

    3) How do alternations like ho and hraf show that their unstressed vowels must have been peripheral? I'd argue that ho and hraf are from earlier *ḥār [ħɑr] and *ḥarífa [ħəˈrɪfə]. (Note Walter Crum cites Sahidic ho, not *hor, and no other dialects have preserved word-final -r either.)

    "Do you know Peust's monograph on Egyptian phonology? Might be an interesting read for you."

    Do you mean Egyptian phonology: An introduction to the phonology of a dead language? Yes, it's very interesting although I'm skeptical of his claim that ayin (ˁ) was pronounced as a /d/ right into the Middle Kingdom as he writes in Section 3.1.

    ReplyDelete
  3. 1) Yes, personally I think that's quite likely, and there are clear cases where all vowels are attested for the same word.

    (I forgot to mention the evidence from Coptic that Peust adduces in § 5.8.3, which is very indirect, though.)

    2) The claim that there is a correlation is from Schenkel 1983 (Aus der Arbeit an einer Konkordanz... II: Zur Pluralbildung des Ägyptischen, the reference is given in Loprieno's book), whose reconstructions are based mainly on the work of Osing (Die Nominalbildung des Ägyptischen, Mainz 1976), who reconstructed several hundreds of nominal forms. The reconstruction of unstressed vowels is based on Akkadian (and a few Meroitic) transcriptions, accent alternations and (most speculative of all) the assumption that the derived nouns fell into clearly defined patterns as in Semitic (e.g., *CáCiC for active participles of triradical verbs), which is used to assign unaccented vowels based on other words of the same pattern. I don't have Schenkel's book here to give you a specific example.

    3) Right, it's ho, and it is a bad example, unless you work in Peust's syllabification framework, where you would need *ḥárV, and where it would make sense to assume that the unstressed vowel was the i of *ḥarifa. Otherwise, all the examples I can think of only support unstressed *a. I collected all the pronominal state examples I could find a while ago, but I don't have them here now, and I don't remember whether there was any evidence for unstressed i/u.

    (By the way, why do you reconstruct ḥarífa, but not ḥāra?)

    About Peust: Yes. Well, yes, the ayin thing is controversial and I'm also skeptical.

    ReplyDelete
  4. "(I forgot to mention the evidence from Coptic that Peust adduces in § 5.8.3, which is very indirect, though.)"

    My position, as far as I'm aware, doesn't conflict with what Peust says in 5.8.3 regarding Coptic. Word-initial *y- is omitted by 1500 BCE, just as in Minoan by areal influence. So the same development of Old Egyptian *a- and *ya- reflects this early merger.

    Peust's six classes here correlate with what I understand to be *tap 'head' (later corrupted to ape by misinterpretation of its t-), *awina 'colour', *abāṭa 'month', *alācat 'milk', *asāru 'sheep' and *wahāra 'dog'. Unstressed word-initial *a- of Middle Egyptian (from either Old Egyptian *ʔa- or *ya-) is understandably preserved as is in Coptic while the lowering of a vowel next to a pharyngeal is expected. The reduction of *wa- [wə] to *u- as in *wahāra makes good sense too. I take it that Peust recommends *i- in 'sheep' but I would like to know if he provides examples of contrasts then between his *is- and *as-. I challenge by asking: Are we sure that /s/ hasn't affected the expression of the preceding vowel by raising it?

    "The reconstruction of unstressed vowels is based on Akkadian (and a few Meroitic) transcriptions, accent alternations and (most speculative of all) the assumption that the derived nouns fell into clearly defined patterns as in Semitic [...]"

    Yes, the danger is that if the Old Egyptian vowel system has underwent radical changes through vowel reduction as I explain it, then the Egyptian vowel system mirrors the Proto-Semitic system only superficially. Then there are some notable differences in stress placement that should remind us to be cautious about this.

    "(By the way, why do you reconstruct ḥarífa, but not ḥāra?)"

    I'm not sure what you mean. I would pronounce "his face" as *ḥarífa [ħəˈrɪfə] with stress on the second syllable. That stress placement must be a carry over from Proto-Egyptian when the oblique form in *-i was still in effect (ie. *ḥar-í=fa = 'face-GEN-3ps). The stress has shifted according to the penultimate rule, causing the reduction of the now-unstressed long a to an unstressed schwa. In this way, there are still remnants of the Proto-Egyptian case system. The citation form however would be *ḥār with a long vowel whose length becomes a matter of qualitative contrast as it becomes [+back].

    As we can see, this can get complicated and I think of Old and Middle Egyptian as being rich in obscure vowel alternations that are hold-overs from bygone times.

    ReplyDelete
  5. (Actually come to think of it, I wonder if the *a being commonly reconstructed for "face" is not a short *u (according to my system at least), hence *ḥur then (but *ḥarífa would remain the same). A long vowel in this closed syllable has been disturbing me but I wonder if there's some external evidence for that vowel quality.)

    ReplyDelete
    Replies
    1. The -i in harif(a) I believe was actually a 'u', James P Allen mentioned this in his newest book. And if you notice how -i:(j) reflects adjectival nisbas and -i: indicates the first person suffix pronoun then you'd simply either have -a: or -u: left.
      In Modern Arabic there is also a voluntary adding of -u, (-i or -a, depends on dialects) before suffix pronouns when certain verbs preceded the suffix pronoun.
      In Beja (one of the closest Afroasiatic Cushitic languages within Egypt who have another native tongue other than Arabic) has strikingly similar suffix pronouns to Egyptian, for example,
      -f (he) = - ooh (Beja Cushitic language) *for some reason Afro-Asiatic languages have - h- and Egyptian changed it to -f-, probably a similar situation with Latin -f- words turning into "silent" H in Spanish (figlio - son in Italian, hijo - son in Spanish) **

      Delete
    2. I'll have to track down Allen's point of view. I've been going by Loprieno's reconstruction in Ancient Egyptian: A linguistic introduction (1998) where he proposes *ḥar-i-f and explicitly mentions the intervening genitive suffix in *-i, a case suffix securely shared by Semitic languages.

      As for the 3ps.masc. suffix -f, I've presumed that it derives from earlier **-hu in which case the sound change has gone the other way: labialized [hʷ] > [f]. Likewise a 3ps.fem. **-hi would spirantize to -s by Old Egyptian (palatal [hʲ] > [s]) in which case there's alignment with the Semitic pronouns (Akkadian šū 'he' and šī 'she'; Hebrew 'he' and 'she'). There seems to be a connection there somehow but now we're delving into the mysteries of Pre-Egyptian.

      Delete
    3. Allen (in Ancient Egyptian 2013 p. 61) reconstructs ḥr.f as ḥarúf because of the variant Coptic cognates hrεf/hraf "his face" (imagine the Coptic alphabet here), since the ε/a vowel alternation was common for original stressed u. Analogously, he has hr.Tn = *harúTun > Coptic hrηtn where the e-quality η is from *ú.

      This also correlates with the concept of -u having been an early Egyptian masculine ending lost in most masculine nouns, save those with a lingering -w, to apocope.

      Delete
    4. Fine. So what about the lack of "u" and seeming presence of "i" before the possessive ending fossilized in the name Bukurninip (= bȝk-n-rn-f) then?

      Delete
    5. We shouldn't solely go off of Cuneiform renditions alone because the scribes butchered Egyptian words, for example:

      nfr khpr re=
      napkhururia
      naphkhororia
      napkhuru-ri(ya)
      napkhuria

      nb maat ra =
      annumuria
      nimmuaria
      nibmuaria

      Also taking into account dialectal variations as well as errors in pronunciation.

      I am also curious to know how some Egyptologists have come up with Coptic a/e variation = Old Egyptian u ...since supposedly according to Old Bohairic pronunciation e = ae sound, and Old Bohairic Coptic a = a in father ... unless Koine Greek influenced/altered the original Egyptian pronunciation of e/a how does it go back to an original u? Same concept is applied to Old Bohairic Coptic 'H', they say sometimes it goes back to an original 'u' but in Coptic 'H' has either an 'a/ay' sound or an 'i' sound?! I just can't seem to comprehend how a 'u' can turn into an 'a' so unpredictably or maybe I'm just not understanding something.


      Delete
  6. My position, as far as I'm aware, doesn't conflict with what Peust says in 5.8.3 regarding Coptic.

    Okay, I don't have Peust 1995 here to check whether the contrast can be easily attributed to the environment.

    I'm not sure what you mean., and A long vowel in this closed syllable has been disturbing me but I wonder if there's some external evidence for that vowel quality.

    The reconstructions of Osing, Loprieno, Peust (among others) assume that Coptic o and a originate from
    the same type of syllable (both open or both closed), and both originate from short vowels, so the standard reconstruction would be *ḥaríf and *ḥár, not *ḥá:r. Is there a reason for giving up this hypothesis?

    I think of Old and Middle Egyptian as being rich in obscure vowel alternations that are hold-overs from bygone times.

    That's quite plausible, but the standard reconstruction contains very few obscure alternations. Do you have evidence for specific alternations (other than those which would have arisen from the assumed reduction of unstressed vowels)?

    ReplyDelete
  7. The citation form however would be *ḥār with a long vowel whose length becomes a matter of qualitative contrast as it becomes [+back]

    Oh, do you would assume that only long a became o, while short unstressed a became e/a/nothing. But since you give *a as [ə], you can reconstruct stressed short /a/ and a separate unstressed phoneme /ə/ (unless that could be dealt with by syllabification rule)

    ReplyDelete
  8. "Oh, do you would assume that only long a became o, while short unstressed a became e/a/nothing."

    Yes. So Middle Egyptian *anāka "I" becomes Sahidic Coptic anok. If a long vowel (ie. stressed open vowel) is put there to explain Sahidic o, it also nicely correlates with Proto-Semitic *anāku, also with long vowel. Around 1500, long must have been still [-round, +back] while the short stressed *a was to the front. Unstressed *a had been reduced to mid-central schwa.

    It was around 1100 that long turned to *o, precisely at the same time as the neighbouring Canaanite Shift.

    "But since you give *a as [ə], you can reconstruct stressed short /a/ and a separate unstressed phoneme /ə/"

    No, I take this schwa (unstressed *a) to be the reduction of stressed *a, *i and *u. I choose necessary simplicity as the soundest methodology unless there is evidence to contradict my idea. What I would like to see is evidence clearly showing vowel contrasts in unstressed syllables.

    "That's quite plausible, but the standard reconstruction contains very few obscure alternations."

    I wouldn't go that far. Egyptian had a few broken plurals like nṯr 'god' whose vowel alternations are preserved in Coptic between the singular noute and plural nter going back to *nācar and *nacūru respectively. Once can presume an earlier Proto-Egyptian stage where the alternations were less obscure: *nācur-u vs. *nacūr-aw. Note that Egyptian shares broken plurals with South Semitic languages like Arabic, implying areal influence.

    ReplyDelete
  9. it also nicely correlates with Proto-Semitic *anāku, also with long vowel

    Berber (Tuareg năkk) supports a short vowel, though.
    Do you reconstruct every stressed *a as long, or do you assume Coptic ō goes back to short stressed *a? If not, what is the source of Coptic ō in your reconstruction?

    ReplyDelete
  10. What I would like to see is evidence clearly showing vowel contrasts in unstressed syllables.

    Well, I think there is no crystal-clear direct evidence as there is always room for doubt in the contemporary transcriptions. But before abandoning the traditional reconstruction entirely, I would check whether the correlations (in the occurence of written -w, Peust's classes for initial vowels, and the fact that the Akkadian transcriptions of unstressed vowels are claimed to support the reconstruction of vocalization patterns associated with certain types of derivations, like *CáCiC rather than *CáCaC, *CáCuC for active participles) can really be plausibly attributed to other phonological factors or even to chance. If another explanation (e.g., the surrounding consonants) can be found for the data given by Osing and Schenkel, your reconstruction would clearly be superior. If not, a contrast of unstressed vowels would seem to be quite an economic explanation to me.

    ReplyDelete
  11. Once can presume an earlier Proto-Egyptian stage where the alternations were less obscure: *nācur-u vs. *nacūr-aw.

    Interesting, alternations of this type would clearly support your reconstruction. But I'm a bit skeptical, since the only reason (in the standard theory) to reconstruct *a in *nācar seems to be *nVcárVt "goddess". Under your Proto-Egyptian reconstruction, the paradigm would again be irregular, having both *nacār-at-u and *nācur-u. Broken plurals might well go back to Proto-Afroasiatic, as they appear in all branches except Omotic.

    ReplyDelete
  12. "Berber (Tuareg năkk) supports a short vowel, though."

    Before accepting this evidence, could you first explain the source of the gemination in -kk?

    "Do you reconstruct every stressed *a as long,[...]"

    No. Stressed *a remained a in Sahidic (eg. *waˁ(i) 'one' > oua).

    However stressed first became *o wholesale in Late Egyptian, via the ill-named Canaanite Shift that affected both NW Semitic and Egyptian languages simultaneously. Then *o either remained o in closed syllables (eg. anok 'I', onḫ 'life') or ō in open syllables (eg. rōme 'people').

    "If not, a contrast of unstressed vowels would seem to be quite an economic explanation to me."

    Let me challenge you. How does one convincingly explain the pattern of stress accent and vowel selection seen in situations of a noun plus possessive suffix? As per Blazek, for example, why *ḥaríf 'his face' (with accent on the suffix), and yet *nībif 'his lord' (with a lengthened vowel from *nib), and yet further still *rinn-if 'his name' (with double consonant and no indication of stress, from *rin)?

    I charge that this is all ad hoc and given the same evidence, everything points more strictly to: *ḥar=ífa 'his face' (< *ḥur), *nīb=afa 'his lord' (< *nība ~ *nib) and *ran=ífa (< *rin). The stress then predictably lies on the penultimate, unless drawn to a long vowel present in the root noun.

    To add further evidence in my favour, likewise *ˁānaḫ 'life' plus possessive *-sa becomes *ˁānḫ=asa with stress on the first syllable with long vowel, confirming both this rule of accent placement here as well as the fact that there was a long vowel in *ˁānaḫ... which becomes Late Egyptian *onḫ (Canaanite Shift) and then remains as such in Sahidic because *o is in a closed syllable. Since *ˁānḫasa is directly reflected in the cuneiform rendering, Anhasanpiyati 'Her life [is] of the Aten' (nb. -asa- is reflected in the possessive suffix), everything is tightly accounted for, both the vowels and stress pattern.

    "Under your Proto-Egyptian reconstruction, the paradigm would again be irregular, having both *nacār-at-u and *nācur-u. Broken plurals might well go back to Proto-Afroasiatic, as they appear in all branches except Omotic."

    I'm open to debate on broken plurals because I have no certain answer. However you're unfairly charging me to explain what others as far as I'm aware cannot.

    Loprieno's alternative involving singular *nācar beside plural *nacúr-w is no more self-explanatory and testifies against your claim that "the standard reconstruction contains very few obscure alternations". This obscure alternation demands to be explained one way or another.

    It's quite possible though that it's *nacārat 'goddess' instead that preserves the second vowel of a long-gone Proto-Egyptian stage. Then again, that still doesn't explain the source of Loprieno's infixal *u (nor my alternative with ). Does anyone else's model explain it and, if so, how?

    ReplyDelete
    Replies
    1. I've done self-research on broken Plurals myself, and for some reason my gut said Egyptian did not contain broken plurals. In Coptic a lot of the differences between vowel quality has to do with the movement of stress, from what the original vowel would have been. It appears that the plural marker was redundant and people stopped using it replacing the plural marker by using the plural definite article for most words.

      I also don't feel like the Egyptians wouldn't have continued to use hieroglyphs without vowels to indicate a difference in vocalisation of the plurals if only -w was needed to express plural then why have broken plurals along with the plural marker?

      This also didn't carry over into Coptic, no indication of broken plurals other than the changing of vowels due to movement of stress accent.

      With "nácar" the stress moves a syllable over to the right "nacárwa" or "nacáraw" and this stressed long "a" may have had an alternative pronunciation into the creation of Coptic. Wouldn't a broken plural be more like "nácar = nicárwa or yancúrwa/incúrwa"?

      Delete
  13. [...]could you first explain the source of the gemination in -kk?

    Good point, I can't. Maybe a long vowel was the cause, but it would be good to have more examples with gemination and an Afroasiatic etymology (Prasse attributes gemination in verbs to *h).

    How does one convincingly explain the pattern of stress accent and vowel selection seen in situations of a noun plus possessive suffix?

    I don't know. Nonetheless, some words on two reconstructions. In the standard system (by this I mean Osing, Schenkel, Loprieno), we get:

    nb=f: absolute níb-Vw, pronominal níb-wVf (the w is attested in writing). The change in syllable structure is predicted by the two-syllable-rule. Why do you reconstruct ī rather than i? Is it because of Coptic ē, while i would be derived from short i? Could you maybe give a small table of your reconstruction of the basic developments? That would be very helpful for understanding your ideas.

    rn=f: absolute rín, pronominal rín-if. Blazek's geminated n seems to be unfounded (I don't have Vycichls work here), and the accent in the pronominal form wass on the first syllable (Coptic rimf).

    As we see, the stress is unpredictable under the standard system. But I'm skeptical that your system makes it any more predictable, as it does not explain why we have rín-if instead of rin-íf.

    No. Stressed *a remained a in Sahidic (eg. *waˁ(i) 'one' > oua).
    [...]that there was a long vowel in *ˁānaḫ... which becomes Late Egyptian *onḫ (Canaanite Shift)[...]

    The idea that only long á became o/ō is interesting and quite appealing, given the situation in Canaanite. Do you have cuneiform evidence for the change short á > Coptic a (preferably without a neighboring pharyngeal as in "one", as they tend to change vowels to /a/ in Semitic)?

    ReplyDelete
  14. Since *ˁānḫasa is directly reflected in the cuneiform rendering, Anhasanpiyati 'Her life [is] of the Aten' (nb. -asa- is reflected in the possessive suffix), everything is tightly accounted for, both the vowels and stress pattern.

    Well, I agree with your reconstruction of the stress pattern, but the rendering does not show that the stressed *a was long. (Just as a side remark, the rendering does not really lend independent support to more than *ˁanḫVsV, as you also reject unstressed i/u in other renderings and there is no reason for preferring *a as the only unstressed vowel other than the economy of your proposed system.)

    Does anyone else's model explain it and, if so, how?

    I don't think so.

    This obscure alternation demands to be explained one way or another.

    Well, my explanation would be a CaCuC broken plural pattern. We don't have an explanation for the broken plurals in other branches either, except for reconstructing some of them (especially those with -a-) for Proto-Afroasiatic. Afroasiatic languages just have them, it seems quite natural to assume that Egyptian preserved some of them.

    However you're unfairly charging me to explain what others as far as I'm aware cannot

    No, I just wanted to point out that CaCuC seems to be obscure under any known reconstruction :-)

    Don't get me wrong, I totally agree that the standard reconstruction is speculative and its results are often not quite satisfying. Your analysis is very interesting and obviously more economic than the standard reconstruction, but I'm not convinced that it explains as much as the standard reconstruction. In particular, it is not clear to me that your reconstruction can explain all the evidence for unstressed i/u and correctly predict the knwon pairs of Cuneiform and Coptic forms (well, at least a higher percentage of pairs than the standard reconstruction). Proving this convincingly would probably require looking at the several hundreds of words collected in the studies arguing for the more standard reconstructions. If your reconstruction stands up to this test, you have probably revolutionized the reconstruction of Egyptian.

    ReplyDelete
  15. Correction: Prasse attributes gemination to *w, not to *h.

    ReplyDelete
  16. Speaking of corrections, in fear of spreading insidious internet memes, I'm in truth unsure whether Anḫasanpiyati is attested or just a hypothetical concoction of Fecht's. I'll retract this until I can verify. That being said, I still stand behind a reconstruction with *ˁān(a)ḫ=asa with a long vowel in the first syllable, thereby pinning the stress there.

    "Maybe a long vowel was the cause, but it would be good to have more examples with gemination and an Afroasiatic etymology (Prasse attributes gemination in verbs to *w."

    Yet this doesn't apply well here. In many respects Proto-Berber has innovated greatly when compared with both Egyptian and Proto-Semitic. Marijn and I have had a few discussions on this odd gemination but some of it appears to be a vexing mystery for now.

    So I put greater trust in the more conservative Semitic evidence: Akkadian anāku.

    "Why do you reconstruct ī rather than i?"

    To get the accent on the first syllable of *nīb=afa without a two-syllable-rule, otherwise it would be **nab=ífa. The placement of stress is solely the product of a default penultimate stress rule secondarily attracted by length.

    "As we see, the stress is unpredictable under the standard system. But I'm skeptical that your system makes it any more predictable, as it does not explain why we have rín-if instead of rin-íf."

    You seem very assured of the equation of Coptic rin-k 'your name' with Ancient Egyptian rn.k and rim-f 'his name' with rn.f despite rin-t 'my name' (≠ rn.ỉ). The latter form makes it look like these are all just alternative forms along the lines of rnt-k 'your name', a reduction of possessives using ntēi 'of me', ntak 'of you', ntaf 'of him', etc.

    By contrast, note how ho 'face' isn't marked with -t in the 1ps possessive form hra-i 'my face' which relates better than 'name' to its earlier Egyptian form, ḥr-ỉ.

    Without Coptic evidence, we're left with the equation of Babylonian Bukurninip with bȝk-n-rn-f 'servant of his name' which could just as well reflect, say, *Boˀka-n-raníf.

    "Do you have cuneiform evidence for the change short á > Coptic a (preferably without a neighboring pharyngeal as in 'one', as they tend to change vowels to /a/ in Semitic)?"

    Okay, let's talk about Egyptian zj 'man' > Coptic sa in all dialects.

    "Well, my explanation would be a CaCuC broken plural pattern."

    And that's a decent explanation, I think. For all we know, such a pattern represents a most ancient Afro-Asiatic plural form only later marked with an explicit plural in *-aw.

    ReplyDelete
  17. The latter form makes it look like these are all just alternative forms along the lines of rnt-k 'your name', a reduction of possessives using ntēi 'of me', ntak 'of you', ntaf 'of him', etc.

    Oh I wasn't aware of that problem, I just took the Coptic form from the Erman/Grapow dictionary. So my counterexample collapses. But I think there is another problem:

    To get the accent on the first syllable of *nīb=afa without a two-syllable-rule, otherwise it would be **nab=ífa

    "Face" must be *ḥār(a?) in your system to explain Coptic ho, but then we would expect *ḥár-VfV, Coptic *hōr= rather than hra=. So either the idea that *ӑ did not become o/ō or the accent rule seems to be incorrect.

    Okay, let's talk about Egyptian zj 'man' > Coptic sa in all dialects.

    I would reconstruct *zíj. Is there any evidence for *zaj? The internal evidence of Coptic suggests that a is just the correlate of ē in closed syllables. There certainly are cases where it is derived from *a ("seven": cuneiform šapḫa > Coptic sašf), but the *a is caused by ḫ etc. in the standard system (Peust § 5.6.2.1). Do you have other evidence for *a > a?

    ReplyDelete
  18. "'Face' must be *ḥār(a?) in your system to explain Coptic ho, but then we would expect *ḥár-VfV, Coptic *hōr= rather than hra=. So either the idea that *ӑ did not become o/ō or the accent rule seems to be incorrect."

    Yes, I noticed my paradox a few comments ago which is why I've already adapted my reconstruction while we were talking. Notice above that I had revised my reconstruction to *ḥur with short vowel. Without a long vowel to draw away from the default penultimate stress, the possessive becomes *ḥar=ífa. Sorry about the confusion; work in progress.

    So this means that Middle Egyptian's short *u, together with , merged to o by the 1st millennium BCE. (This is phonologically reasonable because in general short vowels tend to fall while long vowels tend to rise.)

    Interestingly enough, I've already been reconstructing *kuˀ 'ka, embodiment, soul' to explain the ko- in the Coptic month of Koiak as well as the -gu- in Αίγυπτος 'Egypt' which we know is taken early on from something like *Ḥaˀat-Kuˀ-Pataḥa 'Memphis' (ie. literally 'House of the embodiment of Ptah'). A vowel *a in this particular word, whether long or short, doesn't explain the evidence.

    "I would reconstruct *zíj."

    I thought you might, but then what's the difference between *zíj and *zī in terms of pronunciation? How do those two sequences manage to be kept separate during their development? I presume then that the corresponding feminine form becomes a disyllabic word *zíjat in your system.

    However is it not possible that the form is *za instead (with monosyllabic feminine *zat)? How do you explain Sahidic sxime 'woman' from z(.t) ḥym.t with only s- and no trace of its vowel nor of its glide? From my view, it makes more immediate sense that unstressed *za(t)- was reduced to Coptic s-.

    ReplyDelete
  19. Yes, I noticed my paradox a few comments ago which is why I've already adapted my reconstruction while we were talking.

    Right, sorry, I had forgotten about your comment.

    Interestingly enough, I've already been reconstructing *kuˀ 'ka, embodiment, soul' to explain the ko- in the Coptic month of Koiak as well as the -gu- in Αίγυπτος 'Egypt' which we know is taken early on from something like *Ḥaˀat-Kuˀ-Pataḥa 'Memphis' (ie. literally 'House of the embodiment of Ptah'). A vowel *a in this particular word, whether long or short, doesn't explain the evidence.

    Well, the *u in these compounds might well have been unstressed, and in the standard reconstruction, the rules applying to unstressed vowels are quite different from those applying to stressed vowels. So I still don't see any motivation for *ḥur and *ú > o other than the fact that it saves the stress rule and makes it possible to allow *ā only in open syllables.

    I thought you might, but then what's the difference between *zíj and *zī in terms of pronunciation? How do those two sequences manage to be kept separate during their development?

    That's an interesting question. I actually prefer Peust's system, where one gets *zíjV. If V was *a, there is no problem.

    I presume then that the corresponding feminine form becomes a disyllabic word *zíjat in your system.

    I don't know whether *zj.t is ever attested in writing. If not, I'd reconstruct *zitV, which is compatible with the Coptic form.

    ReplyDelete
  20. "Well, the *u in these compounds might well have been unstressed, and in the standard reconstruction, the rules applying to unstressed vowels are quite different from those applying to stressed vowels."

    You're not just dismissing Greek Αίγυπτος but the added evidence from Byblos reflecting Ḥi-ku-up-ta-aḥ ~ Ḥi-ku-ta-aḥ (EA 84.37 and EA 139.8). I think it takes a lot of effort to deny *u here when there are at least three separate instances of it from foreign testimony.

    "I actually prefer Peust's system, where one gets *zíjV. If V was *a, there is no problem."

    But where does this *a come from then? From a Pre-Egyptian masculine marker? This still sounds ad hoc.

    ReplyDelete
  21. I think it takes a lot of effort to deny *u here when there are at least three separate instances of it from foreign testimony.

    My comment was unclear, sorry, I'm absolutely not denying *u here, I'm claiming that the change from *u to Coptic o only occured in unstressed syllables, as I don't know of evidence for the same change in syllables that we know were stressed. (It seems plausible that Koiak derives from *ku3V-ḥúrV-ku3V)

    But where does this *a come from then? From a Pre-Egyptian masculine marker? This still sounds ad hoc.

    I could imagine that it reflects the absolutive ending. But, yes, it's speculative. However, working in Peust's system, it is possible that it was indeed pronounced as *zī (hence the j was very rarely written, just as w was in *-uw, *-iw). Since Peust assumes that Coptic o,a,e come from open syllables, sa is accounted for.

    ReplyDelete
  22. "My comment was unclear, sorry, I'm absolutely not denying *u here"

    Then this is one instance showing Middle Egyptian stressed short *u becoming Sahidic o, as I had claimed.

    Why do you place the accent on your preposition *ḥúrV? A more natural placement of accent would lie on one of the instances of *kuˀ... but this would force you to accept that accented *u could become Sahidic o and you won't have that.

    Nonetheless, I suggest *Kuˀ-ḥara-kuˀ, the 'Ka upon Ka'. Accent lies naturally on the first syllable of the compound. I would gather this was reduced in Late Egyptian to something like *Koˀḥiyak(ə) > Sahidic Koiak.

    "However, working in Peust's system, it is possible that it was indeed pronounced as *zī (hence the j was very rarely written, just as w was in *-uw, *-iw). Since Peust assumes that Coptic o,a,e come from open syllables, sa is accounted for."

    So you claim that long becomes Sahidic a along with short *i, and then despite the added length it disappears altogether in sxime 'woman'? What about rīˁ 'sun' (Sahidic )? Can you elaborate on the phonetic processes you're suggesting here? I can't make sense of it.

    ReplyDelete
  23. Why do you place the accent on your preposition *ḥúrV? A more natural placement of accent would lie on one of the instances of *kuˀ... but this would force you to accept that accented *u could become Sahidic o and you won't have that.

    Well, apart from perceived naturalness there is no compelling independent evidence for either *kúˀ-... or *...-ḥúrV-.... Before accepting *ú > o I would like to see more evidence than a single example where the placement of stress is uncertain. The standard reconstruction is based on hundreds of word forms. As I said before, you need to check your reconstruction on much more data.

    So you claim that long *ī becomes Sahidic a along with short *i, and then despite the added length it disappears altogether in sxime 'woman'? What about rīˁ 'sun' (Sahidic rē)? Can you elaborate on the phonetic processes you're suggesting here? I can't make sense of it.

    Well, s- in sxime comes from the feminine z.t (z.t-ḥjm.t is attested), which was *zit with short *i. The vowel length in *zī is caused by *-j, which was not present in the feminine form. As the masculine form was written z or zj (Edel, Altaegyptische Grammatik § 66) and the feminine only zt, both the Coptic and the Hieroglyphic evidence are accounted for. On the other hand, it is not clear to me why *za should have been written zj.

    ReplyDelete
  24. "Well, apart from perceived naturalness there is no compelling independent evidence for either *kúˀ-... or *...-ḥúrV-...."

    You still haven't accounted for Gk Αίγυπτος and Bab Ḥikuptaḥ, both pointing straight to a vocalism of *kuˀ.

    "Before accepting *ú > o I would like to see more evidence than a single example where the placement of stress is uncertain."

    Perhaps it'd be best if you accounted for the already-cited evidence first before proceeding with yet further evidence you may have difficulty accounting for. Only objective explanations, not subjective denials, have value to me.

    "On the other hand, it is not clear to me why *za should have been written zj."

    One way or another, you admit that unstressed short vowels *must have* merged to schwa before disappearing in words like sxime since their quality evidently doesn't matter to you. By Occam's Razor, a model that doesn't require contrasts that can't be proven is a superior model. You still can't show that there were contrasts between short vowels in unstressed positions and this is a problem for your comparatively complex view.

    "On the other hand, it is not clear to me why *za should have been written zj."

    Yes, in one alternative spelling, a reed leaf precedes the determinative MAN. The literal interpretation is that this symbol directly represents j but unfortunately Egyptian scribes didn't write in literal ways much of the time.

    Judging by Loprieno's LEgy *re:ʕə 'sun' (Sahidic ), surely your *zī would end up rhyming with it. But Sahidic sa doesn't rhyme. So your idea is still problematic.

    A reconstruction of *ziˀ would be an improvement since the change of *i to Sahidic a is already well proven (*rin 'name' > ran). However this dissuades us from the common literal transcription and, given your *zit (without glottal stop), it'd force us to reexamine the role of word-final glottal stops in monosyllabic words (ie. quite likely word-final glottal stops were non-contrastive in open stressed syllables anyway, as in *kuˀ).

    It seems to me then that either *ziˀ or *zaˀ (assuming contrastive word-final glottal stop for now) are more sensible reconstructions to choose from. In unstressed position though, my model prescribes *za(ˀ)- with *a standing for schwa regardless of either reconstruction.

    ReplyDelete
  25. You still haven't accounted for Gk Αίγυπτος and Bab Ḥikuptaḥ, both pointing straight to a vocalism of *kuˀ.

    Yes I have, the standard reconstruction for Koiak contains -ku3-.

    Yes, in one alternative spelling, a reed leaf precedes the determinative MAN. The literal interpretation is that this symbol directly represents j but unfortunately Egyptian scribes didn't write in literal ways much of the time.

    There is evidence for -j from the Old Kingdom, where every written weak consonant should be taken seriously.

    Perhaps it'd be best if you accounted for the already-cited evidence first before proceeding with yet further evidence you may have difficulty accounting for. Only objective explanations, not subjective denials, have value to me.

    *ku3V-ḥúrV-ku3V seems to be quite a plausible explanation for Koiak, as is *kú3V-ḥurV-kú3V. I'm not denying the a priori possibility of the latter reconstruction, I'm trying to show that Koiak is accounted for by the standard reconstruction.

    You still can't show that there were contrasts between short vowels in unstressed positions and this is a problem for your comparatively complex view.

    As long as you don't show me that all the evidence adduced for unstressed i/u in the standard reconstruction can be explained differently, I think unstressed i/u is a plausible explanation for the evidence. You've only explained away a few examples, but enough evidence remains (see the book by Schenkel I mentioned and Peust's 1995 paper. Furthermore, I don't accept that Akkadian scribes randomly chose a vowel for schwa, but always (with the exception of some final vowels) chose the same vowel for a given word.). You should try to explain them away, e.g. by showing the vowel is always determined by the phonological environment, instead of denying their relevance.

    Judging by Loprieno's LEgy *re:ʕə 'sun' (Sahidic rē), surely your *zī would end up rhyming with it. But Sahidic sa doesn't rhyme. So your idea is still problematic.

    No it's not. 'sun' was rʕw (-w is attested), hence *riʕwV, contrasting clearly with *zī. Both rē and sa are regular. Loprieno's long e is only based on assuming that Coptic ē goes back to a long vowel, which
    is not necessary (as Peust explains).

    A reconstruction of *ziˀ would be an improvement since the change of *i to Sahidic a is already well proven (*rin 'name' > ran).

    There is no evidence that the *i in *rin was short.

    However this dissuades us from the common literal transcription and, given your *zit (without glottal stop), it'd force us to reexamine the role of word-final glottal stops in monosyllabic words (ie. quite likely word-final glottal stops were non-contrastive in open stressed syllables anyway, as in *kuˀ).

    Well, maybe *zī goes back to nominative *ziu/*ziw. Then everything would be as expected.

    ReplyDelete
  26. Correction: the attestation of -j I had in mind is not from the Old Kingdom, I misread the volume number in Edel's grammar. So you're right, the written -j cannot be trusted. So this would make things simpler for both Peust's (*zi > sa as *ríni > ran) and your system (which does not need to explain written -j).

    But the point I'm trying to make is that none of your examples provide compelling evidence against the standard reconstruction of the development of stressed vowels, while I would really *not* be surprised if you managed to show convincingly that all the evidence for unstressed i/u can be explained away.

    ReplyDelete
  27. "Yes I have, the standard reconstruction for Koiak contains -ku3-."

    No. It's coherent to claim that an unaccented *u is preserved as o in Sahidic Koiak on the one hand and yet claim that your unaccented *i in *zit- disappears in sxime in the same dialect.

    "There is evidence for -j from the Old Kingdom, where every written weak consonant should be taken seriously."

    Taken seriously, yes. Taken blindly, no. Egyptian is not a literal writing system in the same way as our alphabetic system. Scribes were encouraged to play with symbols in clever ways. If the reed leaf is for underlying *iˀ, we can see how the same symbol might both be treated as a consonant (glottal stop or semivowel y) and used occasionally for its vowel (especially in foreign terms). So *siˀ would be a decent reconstruction that reflects the REED-LEAF in this word. I think, quite frankly, I've convinced myself that this is the better reconstruction over *saˀ (although REED-LEAF might conceivably represent the glottal stop as it did in word-initial position throughout the 2nd millennium BCE too).

    However since we expect **sī to become Sahidic *sē, we know your claim is simply false.

    "There is no evidence that the *i in *rin was short."

    ?? Now you've derailed off the tracks. Both Loprieno and Woodard reconstruct short *i in *rin.

    In my model, only long ī and break down into diphthongs before merging to Sahidic ē while short *i drops to Sahidic a and short *u drops to Sahidic o.

    ReplyDelete
  28. It's coherent to claim that an unaccented *u is preserved as o in Sahidic Koiak on the one hand and yet claim that your unaccented *i in *zit- disappears in sxime in the same dialect.

    Well, some of the two/three vowels in Koiak must be unstressed. In your reconstruction, unstressed *a is preserved as
    a, while *a in your *zat- disappears. Is that more coherent than my claim?

    However since we expect **sī to become Sahidic *sē, we know your claim is simply false.

    You seem to misunderstand Peust's reconstruction, *i in open syllables becomes a.

    So *siˀ would be a decent reconstruction that reflects the REED-LEAF in this word. I think, quite frankly, I've convinced myself that this is the better reconstruction over *saˀ (although REED-LEAF might conceivably represent the glottal stop as it did in word-initial position throughout the 2nd millennium BCE too).

    I can agree with that. So you agree that we have no conclusive evidence for *a > a?

    ?? Now you've derailed off the tracks. Both Loprieno and Woodard reconstruct short *i in *rin.

    What? The reason they reconstruct short *i is that conventionally open syllables are
    identified with short vowels because their reflexes ē,ō,i were thought to be long. There
    is no other evidence for doing so. You yourself ignore the reconstructed quantity when you
    say that o goes back to *ā, where Loprieno and others have *a.

    ReplyDelete
  29. In my model, only long ī and *ū break down into diphthongs before merging to Sahidic ē while short *i drops to Sahidic a and short *u drops to Sahidic o.

    Thanks for the overview. That is not implausble, but seems to be too simple, in particular you need *i > i (e.g., in "nine") and *u has more reflexes (Peust p. 224, he seems to think that the o in koiahk is stressed, but mentions elsewhere that no other example for *u > o is known).

    ReplyDelete
  30. "In your reconstruction, unstressed *a is preserved as a, while *a in your *zat- disappears. Is that more coherent than my claim?"

    Yes. Once the compound *Kuˀ-ḥara-kuˀ erodes to *Kóˀḥiyakə, it's self-explanatory how a in the second syllable simply has no choice but to be preserved lest we utterly mangle the name beyond all proportion. However we can see that the effects of the deletion of the first unstressed vowel in sxime is relatively minor by comparison.

    It's intuitive: Deletion of *a (your unstressed short *a, *i and *u) is the default unless other relevant factors mitigate.

    "You seem to misunderstand Peust's reconstruction, *i in open syllables becomes a."

    That's because my model is a twist on Loprieno's system instead. Fine, so how do you resolve Sahidic nib and Assyutic nep in your Peust-derived system again?

    "You yourself ignore the reconstructed quantity when you say that o goes back to *ā, where Loprieno and others have *a."

    Yes, but Semitic supports long in the 1ps pronoun *anāku which is then related, in my model at least, to Egyptian *anāka. The change of specifically LONG to *o precisely mirrors the Canaanite Shift as I said. You would have me ignore all linguistic context, both prehistoric and historic, just to blindly follow Peust.

    "That is not implausible, but seems to be too simple, in particular you need *i > i (e.g., in 'nine')"

    No. Since my system is derived largely from Loprieno, I have long in *pasīḏa (> LEgy *psit), justified by Babylonian inscription EA 368 which records pi-ši-iṭ.

    "(Peust p. 224, he seems to think that the o in koiahk is stressed, but mentions elsewhere that no other example for *u > o is known)"

    But if Peust thinks, as you say, that stress falls on the first syllable of Koiak, well, I rest my case! It doesn't add up (much like his account of ayin being pronounced as *d throughout historical times).

    I stress the case of Αίγυπτος (backed doubly by Ḥikuptaḥ) since it is undeniable that it's an ancient loan, likely even through a Minoan intermediary *Aikupita. The word for 'ka' is thus without a doubt *kuˀ by all historical accounts since the early 2nd millennium BCE. So we just can't wave away the example of Koiak, showing that indeed Middle Egyptian *u has UNDENIABLY become Sahidic o.

    So you should worry less about the alleged rarity of *u to Sahidic o and focus more on why Peust's theory, or your application of it, fails to explain the historical evidence.

    But there's also *ḫapúri 'beetle' (ḫprỉ, the 'becoming one'). The Greek Magical Papyrus reflects χφυρις (chphuris) in late times. Hesychios claimed as well that the Etruscan word for Greek κάνθαρος 'beetle' was βύρρος (burros). Since scarabs are found not only in Etruscan tombs but at early Minoan sites too, I suspect an Aegean loan from Egyptian by 2000 BCE leading to Etruscan *pur and Minoan *apúri.

    No matter how we interpret it though, the case of the 'beetle' and that of 'soul' are two instances showing how the Egyptian short vowel *u was fairly stable over the millennia, in contradiction to those that claim that short *u > LEgy *e > a like short *i does (as per Loprieno) or some other tall tale of drastic vowel change. Instead the path of *u must be minor (*u > Sahidic o), thus I'd expect *ḫapúri would become Sahidic *xpōre if I've calculated correctly.

    ReplyDelete
  31. "You seem to misunderstand Peust's reconstruction, *i in open syllables becomes a."

    Just to add, on the surface alone, it's not even phonologically plausible that a long high vowel in your *zī 'man', in an open syllable no less, *descends* to a low vowel like a. In countless languages around the world, the commonmost pattern of evolution is that of long vowels rising and *short* vowels falling over time. So I'll assume you're depending on diphthongization and the spontaneous omission of the glide to make this change happen. Improbable.

    ReplyDelete
  32. Once the compound *Kuˀ-ḥara-kuˀ erodes to *Kóˀḥiyakə, it's self-explanatory how a in the second syllable simply has no choice but to be preserved lest we utterly mangle the name beyond all proportion.

    Okay, I accept that this is an argument in favor of your stress pattern. So I accept that stressed *u could also become o.

    [...]so how do you resolve Sahidic nib and Assyutic nep in your Peust-derived system again?

    *níbwV

    No. Since my system is derived largely from Loprieno, I have long *ī in *pasīḏa (> LEgy *psit), justified by Babylonian inscription EA 368 which records pi-ši-iṭ.

    You said before that *ī, *ū > ē, but we have Coptic psit. You would get *psēt.

    So we just can't wave away the example of Koiak, showing that indeed Middle Egyptian *u has UNDENIABLY become Sahidic o.

    ??? I said twice that I reconstruct ku3V. My objections were only about stressed *ú > o, not the possibility of *u > o in general.

    Instead the path of *u must be minor (*u > Sahidic o), thus I'd expect *ḫapúri would become Sahidic *xpōre if I've calculated correctly.

    You're ignoring the examples where Cuneiform u became e or a (see Peust). The situation must be a bit more complex.

    ReplyDelete
  33. "You said before that *ī, *ū > ē, but we have Coptic psit. You would get *psēt."

    Yes, and diphthongize, lose their respective glides, and then merge together to . Overall, only short *a or *i reflect Sahidic a while short *u merges with the o from .

    Back to this merged from MEgy and , sometimes i appears as in tiou 'five' (nb. Callender *dīyaw) and psis 'nine' which I presume is caused by this earlier lying in an *open* syllable (ie. it's free to lengthen and rise in such an environment).

    Note how the bifurcation of merged and to or *i just before the Coptic era mirrors so poetically the similar bifurcation of to *o or that I've already described above and which is historically backed by the Canaanite Shift.

    "??? I said twice that I reconstruct ku3V."

    But you've also said more than once that the first *kuˀ in Koiak was UNACCENTED which violates your other rule that the unstressed vowel should disappear as it does in sxime. I take it your incoherent rules are simply in a state of transition.

    ReplyDelete
  34. Marijn and I have had a few discussions on this odd gemination but some of it appears to be a vexing mystery for now.

    I think it's useful to add a more direct account of our discussion here.

    In Proto-Berber there's no internal evidence at all for shortening of long a, nor reason to think there is a secondary nature of the geminate, the form can reliably be reconstructed as: *năḱḱ, Kossmann (1999) shows that the palatovelars should be considered separate phonemes in Proto-Berber. But with the exciting recent research on the recently discovered Proto-Berber , the data is going to have to be re-examined. A striking amount of nouns that seem to have Palatovelars, have a in the root. But this is going to be researched more in-depth, and I'm going to have make sure this 'gutfeeling' I have about the glottal stop in these words, is numerically relevant. The word for 'I', has no such glottal stop though, so the palatal value of the final geminate (and the geminate itself) remain a mystery.

    ReplyDelete
  35. I think it's useful to add a more direct account of our discussion here.
    Thank you! Yes, the development of what is k in the other branches of AA in Berber seems to quite obscure, with *năḱḱ, the -ɣ in the first person singular and the -k in second person pronouns all corresponding to -k- with some vowels. Vowel length is also quite obscure, I don't know any cognate set in which vowel length in Berber clearly correlates with vowel length in the proto-language of another branche.

    ReplyDelete
  36. What can these palatovelars in *năḱḱ tell us about the vocalism of Egyptian ỉnk if the source of the palatalization can't even be explained. It could be from anywhere, although more likely to be something unique to Berber itself. From what I can see, Berber's vowel system is less conservative than both Semitic and Egyptian so I doubt the archaicism of both the palatalization and simultaneous gemination.

    ReplyDelete
  37. @PhoeniX: I've just stumbled upon Zenaga ni'kʰ- (Kossmann 1999, 180). Does this mean that we need a glottal stop in Proto-Berber here, or is the situation more complicated? (And why does it have an i?)

    ReplyDelete
  38. Interesting all the more since Lameen Souag at his blog Jabal Al-Lughat says "Zenaga /ʔ/ regularly corresponds to pan-Berber ."

    ReplyDelete
  39. Haha, man, I can't believe I didn't check that (whenever I reconstruct ANYTHING for Proto-Berber, the first thing I do is check Zénaga).

    "Interesting all the more since Lameen Souag at his blog Jabal Al-Lughat says "Zenaga /ʔ/ regularly corresponds to pan-Berber *ɣ."

    Zng. /ʔ/ has a double origin, and .

    "Does this mean that we need a glottal stop in Proto-Berber here, or is the situation more complicated? (And why does it have an i?)"

    And yes, this glottal stop means there is very little reason to not reconstruct it. (There isn't reason to think that Zénaga is particularly innovative with glottal stops, if, at all) Glottal stops sometimes disappear or move for quite obviously phontactical reasons, a root like *nɣʔ 'to kill' would become *nʔʔ, which resulted in a Zng. root *ʔnʔ.

    As for the i, I’ll have to explain a bit about the Zénaga vowel system. Zénaga lost the distinction between long and short vowels a long time ago, thus *i and *u merged with ə and *a merged with ă. There is some minimal indication that *i and *u might still be very very slightly phonemic, but it will require more research. For the matter at hand it is quite fine to assume a two vowel system ə and ă. A secondary vowel length developed by compensatory lengthening of Proto-Berber *v (Kossmann's *H/β).

    Now this "i". There's a lot of complex phonotactic rules (which I don't pretend to fully understand), but the bottom line is [i] stands for /ə/. Therefore, Zénaga points to *nəʔḱḱ.

    This is great in one way, and annoying in the other. Perhaps is our lost Proto-Afro-Asiatic glottal stop, although, in that case it would have undergone metathesis, and I don't see any good motivation why it did. On the other hand we have in the root which brings us even further away from home than the already problematic .

    And to close this post, I'll just supply the form that Taine-Cheikh 2008/2010 gives for the 1sg. pronoun in Zénaga. Taine-Cheikh is a very good researcher, and has produced two fantastic dictionaries supplied with (sometimes questionable, but mostly solid) etymological notes. Therefore these dictionaries, along with Kossmann 1999 have become incredibly important resources for Proto-Berber research. She gives: niʔK where capital K stands for a geminate, not that different from the original Zénaga Grammar Kossmann 1999 cites from.

    ReplyDelete
  40. The notion that a complex word-final sequence ʔḱḱ could be maintained over millennia despite the fact that contrasts are the least salient in that position is implausible. The real reason for Zenaga's glottal stop must lie in Zenaga herself.

    ReplyDelete
  41. Very interesting! I've had similar concerns as Glen and wondered whether the Zenaga form might derive from *năɣḱ, where ă became ə in Zenaga because of the palatal ḱ, and where ɣ was assimilated to ḱ in the other languages. But since you say Zenaga has a geminate, the reconstruction doesn't seem so appealing any more. I also admit that *năɣḱ still doesn't really look like 'anaaku.

    ReplyDelete
  42. Michael, this **năɣḱ is purely your own whimsy and the word-final is clearly absurd phonemically. As I said, I object to so many articulatory features being boxed into a position that should be least salient in a word! Let's not indulge further. I cited Lameen's input to raise a point about how we can't just naively reconstruct a Proto-Berber glottal stop whenever it exists in Zenaga.

    Perhaps my memory fails me but doesn't Kossmann reconstruct *nǎḱḱ without glottal stop? This seems just about right. I do still question the palatalization which I can only suspect is post-Berber, presumably to explain Shenwa nəč. If we can rid ourselves of the palatalization in the protolanguage, we'd arrive at a more pronounceable form *nǎkk.

    And Michael, you're going to have to accept that *no* Berber word looks much like their Egyptian or Semitic counterparts because the branch must have innovated considerably on the vowel system and phonotactics. Such an observation says nothing.

    ReplyDelete
  43. Erh, I suppose that should be *nəkk technically. (Kosmann & Dolgopolsky are mixing me up.)

    ReplyDelete