20 Nov 2008

Back to business: emphatic particles and verbal extensions

Now, returning to the safer topic of comparative linguistics, I still am trying to account for how my new solution concerning the prehistoric genesis of Proto-Indo-European's uvular sounds helps (or maybe hinders?) my attempts at trying to figure out the origins of the emphatic particle *[ǵ/g](ʰ)[e/o] which is thus far so hideously reconstructed by current Indo-Europeanists.

However, if we take the velar contained in the nominative first person singular pronoun, which appears to contain the fossilized remnants of the earliest form of the emphatic particle when the pronoun was first coined in the Late Period (ie. *h₁eǵoh₂, literally "(as for) my being here" from *h₁e "here" + *ǵe [emphatic] + *-oh₂ [old 1ps subjunctive]; parallel in development to the 1ps pronouns of Inuktitut uva-ŋa, Aleut ti-ŋ, and Proto-Semitic *an-āku), we are pointed to *ǵe as the most appropriate reconstruction. Any other forms of this particle would then have developed later after presumably being influenced by or merging with other existing words or particles with similar phonetics and meaning. In the earlier Mid IE (MIE) stage, we could then posit an emphatic particle *g̰a derived from Indo-Aegean *k’ə. From there, if comparable to Uralic emphatic *-ka attached to some pronominal stems[1], we might finally reconstruct a Proto-Steppe emphatic particle *k’ə to account for both the Indo-European and Uralic forms. Can you all swallow that? Granted, this all remains tentative for now, but it's worth a shot.

Considering the differing velar stop in the emphatic particle, the verbal extension with uvular stop, seen in PIE verbs like *yeu-g- "to join" whose *g-less counterpart has identical semantics, must not be related afterall as I had previously assumed. Instead I'd like to suggest that it derives from a Mid IE aspectual marker *-ɢ̰a-, which originally might have conveyed a perfective sense. This implies earlier Indo-Aegean *-k’a- (thus Etruscan -ac- [perfective] as in tur-ac-e "was given") and relatable then to the Uralic perfective in *-ka. In this case, a Proto-Steppe perfective suffix *-k’a would be in order to explain the later forms.

That so far is my solution concerning that. Let's see if this idea sticks.


NOTES
[1] Fortescue, Language Relations Across Bering Strait (1998), p.113 (see link) confirms Uralic emphatic *-ka.

22 comments:

  1. I might as well expand on my new theory regarding IE laryngeals. :P

    Based on the lowering properties of uvular consonants, it seems most likely to me that *h2 was a voiceless uvular fricative, i.e. [X]. If you recall, I used to think that it was [x], but that phone is actually easier to pronounce when it is adjacent to a non-low vowel.

    This brings up the question of the value of *h1. At this point, I think it represents at least two different values -- an epenthetic glottal stop before otherwise vowel-initial words, and a voiceless velar fricative (i.e. [x]) elsewhere. The difficulty here, of course, is that some words beginning with *h1 really did begin with [x].

    A bit of internal reconstruction is in order: assuming that the uvularization affected both stops and fricatives, *h1 and *h2 would both go back to an archi-phoneme of [x]. This means that coda* *h1 came from an earlier sequence of *[x] plus non-low vowel, and *h2 came from an earlier sequence of *[x] plus low vowel.

    Additionally, since it seems like coda stops were lenited into fricatives, *h1 could also come from coda *k plus non-low vowel and *h2 from coda *k plus low vowel.

    * Note: it seems to me that IE phonology had syllabification rules which favored open syllables, i.e. most consonant clusters would be analyzed as onsets rather than sequences of coda(s)-plus-onset(s).

    ReplyDelete
  2. Alright, and what evidence do you have from PIE to support this?

    ReplyDelete
  3. I like the idea to use a 'fossilised' form of the emphatic particle, rather than one that is in common use. Because such particles tend to have all kinds of terrible things happen to them like *[ǵ/g](ʰ)[e/o] ;-)

    But you are making a big leap of course to assumed that the 1ps pronoun is indeed a verbification, as you assume it to be.

    If that's true, you have a strong point for the reconstruction of the emphatic. And it's nice to see that other languages would be able to connect with this reconstruction.

    But I'm always a bit scared if reconstructing anything that only has a structure CV; especially such colloquial particles. They could arguable come from anything.

    It's much much to late, I might be rambling a bit. But surely you get my point :D

    I've been wondering whether Sanskrit aham next to tvam is an analogy, or whether it is old.

    Just a quick though.

    if the forms *h1egoH and *tuH both contain the laryngeal (as Beekes reconstructs, thought I don't have a problem with a long vowel *u here) then seeing the alternation of *h1eǵom and *tuom is very similar. Of course the problem remains that we want to have an *ǵʰ or *ǵ to get the Sanskrit h.

    But I just wanted to throw this in, because maybe *tu(H) actually behaves similar to the 1ps pronoun.

    ReplyDelete
  4. Phoenix: "But you are making a big leap of course to assumed that the 1ps pronoun is indeed a verbification, as you assume it to be."

    It should be known that the idea that *h₁eǵoh₂ is analysable in this way isn't even my idea and is already accepted by a few linguists. See Alexis Manaster-Ramer, The Linguist's Linguist: A Collection of Papers in Honour of Alexis Manaster Ramer (2002), p.79

    Additionally I've already went out of my way to show you many examples of the exact same process happening independently in the pronouns of many other languages around the world, particularly (and I repeat!) Inuktitut uva-ŋa "I" and Aleut ti-ŋ "I" which equally attach verbal pronominal markers to their respective deictic stems, uva- and t(x)i-. See Pisani, Studi linguistici in onore di Vittore Pisani, vol.2 (1969), p.552.

    If this is still unconvincing for you, we may add further to the swelling list of these deictic-based formations with examples from Ancient Egyptian *antā-ka "you" (> Sahidic ntok), *antā-fa "he" (> Sahidic ntof) and *antā-sa "she" (> Sahidic ntos); and, Chukchi ɣə-m "I" and ɣə-t "you".

    Compared to other solutions regarding the origins of PIE's 1ps nominative pronoun and suppletion, this is the most explanatory and leaves nothing to mystery. It then follows that this contains a fossilized form of the emphatic particle which then favours *ǵe above all other reconstructions of the particle.

    Phoenix: "if the forms *h1egoH and *tuH both contain the laryngeal (as Beekes reconstructs, thought I don't have a problem with a long vowel *u here)"

    Surely not every pronoun contains a laryngeal! Otherwise we turn the whole pronominal system into a mess of symbols without restraint. The source of the length in the 2ps nominative pronoun (which, mind you, is *optionally* present) must rationally be vocalic, not laryngeal. Reconstructing *Hme "me", *tuH "you (sg.)" and *yuHs "you (pl.)" at the Proto-Indo-European level is, to me, terribly nauseating to look at and seems to go contrary to sensible analysis. I truly believe that any linguist who reconstructs PIE in this manner is missing a few marbles. There, I said it. :P

    ReplyDelete
  5. Oh yes, one more thing concerning those irritating laryngeals I see in some reconstructions of PIE pronouns... I see the same crap going on in Krishnamurti's version of Proto-Dravidian. Take a gander at Krishnamurti, Comparative Dravidian Linguistics (2001), p.336 under the suspicious header "*-H- in Personal Pronouns". Arrgggh! Maddening, isn't it?

    ReplyDelete
  6. I've been sceptical towards reconstructing laryngeals in the pronouns as well. It was 4.30 last night, I didn't feel like thinking over what I was writing :D

    It's just that *tuH seemed to tie in so nicely with *egoH.

    But I guess the Indo-Iranians could've thought the same thing after laryngeals disappeared, and thefore made an analogous tvam to aham.

    I agree that reconstructing the laryngeals is pretty ugly, and assuming a optional lengthening of the *u is probably more plausible.

    A lovely analogy of such a thing is found in my native language Dutch; which has variant forms of most pronouns (though they're all enclitic to the verb)

    ik heb / 'k heb
    jij hebt / je hebt (e is a schwa in auslaut)
    hij heeft / No reduced form. But in questions it does have a reduced form
    heeft hij? / heeftie


    wij hebben / we hebben
    zij hebben / ze hebben

    heb je/hebde? (with -de being a dialectal survival of the old *þu)

    also in the oblique case you have variants:

    Hij slaat mij/me
    Hij slaat jou/je
    Hij slaat hem/'em
    Hij slaat haar/der


    Anyway one more point I liked to make. Last night I said I liked the approach of using a fossilised form of the particle *ǵe to find the 'real' form, since it's less easily subjected to change. But let us not forget that it has been subjected to change in Sanskrit, where it became h which can only come from *ǵʰ.

    So maybe you still end up with a reconstruction of the particle as: *ǵ(ʰ)e ;)

    Which is of course still a lot less ugly than *[ǵ/g](ʰ)[e/o].

    If I remember correctly Beekes' solution to the problem was reconstructing:
    *h₁eǵóH besides *h₁eǵHóm. I find this explanation about as charming as a gorilla. There's no no real reason why you would reconstruct such an ablaut. Even thought it works, it gives more problems than it gives solutions. The question 'what the hell is *-óm, and what is it doing there?'

    A question which is much more elegantly explained with the way you reconstruct it. But then there's the more difficult an irregular shift of Sanskrit h to account for. Sometimes there's no win-win situation.

    But I would definitely place my bets on 'I the verb' rather than 'I with an inexplicable suffix -om in some branches of the language'.

    ReplyDelete
  7. Phoenix: "It's just that *tuH seemed to tie in so nicely with *egoH."

    From what I gather, it started out as *h₁éǵoh₂ and then developed the alternate form, *h₁éǵom, using a different 1ps ending. It's been suggested that, in Sanskrit, *h₁éǵom was reanalysed as an athematic neuter noun *h₁eǵ(ʰ)-óm and assumed to mean "my presence" by folk etymology. From there, *tu-óm "your presence" would arise via analogy with the 1ps. That would explain the accent on the ultima.

    The added aspiration in this 1ps nominative pronoun is probably influenced by a seperate deictic particle *ǵʰo (shown in Jasanoff, Hittite and the Indo-European Verb (2005), p.62, fn.70).

    ReplyDelete
  8. Actually, come to think of it... if *ǵʰo originally meant "back (here), (left) behind" then I might get to reckoning that it's a corruption of o-graded *ǵʰoh₁ < *ǵʰeh₁- "to leave behind". Hmmm, nifty.

    ReplyDelete
  9. Glen: "Alright, and what evidence do you have from PIE to support this?"

    Which part? Or do you mean all of it?

    ReplyDelete
  10. Rob: "Which part? Or do you mean all of it?"

    All parts of your initial quote that happen to be unsubstantiated assertions which are beyond what is commonly accepted by published linguists, of course.

    All your assertions such as the following:
    1. "[x] is actually easier to pronounce when it is adjacent to a non-low vowel" ("easier" how?)
    2. "*h1 represents at least two different values" (based on what data?)
    3. "some words beginning with *h1 really did begin with [x]." (again based on what?)
    4. "Additionally, since it seems like coda stops were lenited into fricatives" (evidence? particulary of Pre-IE *-k becoming a fricative as opposed to a glottal stop?)
    5. "IE phonology had syllabification rules which favored open syllables" (again, why?)

    If you're going to theorize, I except some degree of substantiation, not empty assertions.

    ReplyDelete
  11. Rob, let me just stress, that even though for example German has somewhat a distribution of [x] and [X] near high and low vowels respectively, the more general distribution is [ç] near high, and [x] near low vowels.

    Standard Dutch, including my dialect, have [X] for both g and ch in all contexts. It might be true that languages tend to develop allophony. But you really can't leap upon such an assumption when there's so much counter evidence quite readily available.

    It's funny though, I think I nevertheless agree on the values of the laryngeals; Purely out of a typological point of view. It 'Fits', but I'd call it a suspicion ^^

    ReplyDelete
  12. This pronoun construction scheme looks very interesting. If you excuse an OT moment: reconstructing the 1P pronoun is problematic in Uralistics too, and according to the current consensus, there seem to be two similar but different proto-forms: *mun and *mina. The 2nd could be, however, very transparently analyzed as the demonstrativ root *mi plus the locativ suffix *-na. This feels semantically like more of a stretch than deictic + 1p verbal suffix, but still makes some sense.

    ---

    OK, on with the PIE. Firstly, I'd like to defend Ron's "statement #1". You really only need to pause to think about the actual articulation for a moment. [x] is velar; the vowel articulated closest to that region is [u]. [o] and anything from there on down falls closer to uvular, which you can verify yourself with no other equipment than a mirror required. Featural phonology, too, makes use of exactly [±high] to distinguish velars and uvulars - the same [±high] as found on vowels. (Which is also exactly how it explains lo vowels tending to cause uvularization.)

    Concerning #4 - if you both agree that word-final *H1 is from *k, what does it even matter whether this is [?] or [x]? But that said, [x] however has the plus side that its development can be united with *t > *s / _# under a common rule of final spirantization. This eliminates the AFAIK unexplained asymmetry of *k debuccalizing but *t spirantizing.

    ReplyDelete
  13. Tropylium, your response puts me in a difficult bunch of mindbenders to get myself out of, doesn't it? Thanks! I guess I'll respond to the easier stuff first and chew on the rest for a while...

    Tropylium: "[...] according to the current consensus, there seem to be two similar but different proto-forms: *mun and *mina."

    Yes, I agree that *minä is originally the Uralic 1ps locative/oblique form and that *mun is the original nominative pronoun. What I personally think happened was that the pronominal case system transmogrified due to ergativity. Theoretically minä was already serving in part as an "agentive" in the earliest Uralic stage, and then came to mark the subject as a whole once Proto-Samoyedic started seperating from Proto-Uralic.

    The vocalism of *mun would be affected very easily by 2ps *tun and I believe that Aegean probably reflects the original vocalism (cf. Etruscan nominative mi and oblique mini).

    ReplyDelete
  14. Tropylium: "Firstly, I'd like to defend Ron's 'statement #1'."

    I think you mean "Rob's". At any rate, I think I misunderstood the original statement: "If you recall, I used to think that it was [x], but that phone is actually easier to pronounce when it is adjacent to a non-low vowel." Perhaps I misunderstood what he meant by "that phone". Is he pointing back to [x] or [X]? The text was ambiguous. Mea culpa.

    Tropylium: "This eliminates the AFAIK unexplained asymmetry of *k debuccalizing but *t spirantizing."

    I'm going to have to think about this further because I so far can't think of a convincing counterargument and I have to admit that it does seem to make the theory simpler.

    ReplyDelete
  15. Well, I'll be jiggered. I spoke to soon about Sibilantization of Indo-Aegean *-t and Tropylium's alternate "Word-Final Spirantization" theory ad supra.

    It turns out that Irish English supplies an example of the same type of "assymetrical" softening of word-final stops as I theorize. Word-final English -t in words like what change to fricatives as well. However this only affects alveolar stops (see Hickey, Irish English (2007), p.56), not velar or bilabial stops. Thus it would seem that my theories do indeed occur in practice!

    ReplyDelete
  16. Sure there are examples of single-POA spirantization. *p > f is found in Arabic, *k > x in Mongolian, etc. Likewise for debuccalization (English works again as an example for †t). That's not the asymmetry I protest against, it's that there would be two different lenition processes going on in the exact same environment at the same timeframe.

    And yes, apologies to Rob about the slip of the finger.

    ReplyDelete
  17. Glen: "All your assertions such as the following:
    1. "[x] is actually easier to pronounce when it is adjacent to a non-low vowel" ("easier" how?)"


    "Easier" in the sense that velar obstruents have the articulatory feature [+high] (or at least [-low]), which is shared by non-low vowels.

    "2. "*h1 represents at least two different values" (based on what data?)
    3. "some words beginning with *h1 really did begin with [x]." (again based on what?)"


    These suppositions are not based on IE data. Rather, they are based on the apparent fact that, in languages that lack a phonemic glottal stop, there is nevertheless a phonetic glottal stop which is pronounced, at least in isolation, at the beginning of a word which is phonemically interpreted as being vowel-initial. Despite many (if not most) IEists' claim that all IE words began with a consonant, this fact remains.

    4. "Additionally, since it seems like coda stops were lenited into fricatives" (evidence? particulary of Pre-IE *-k becoming a fricative as opposed to a glottal stop?)"

    For one to obtain such evidence, he must reach outside of IE proper and engage in comparisons with (presumably) related language groups. Uralic offers the best opportunities here.

    As I think you yourself support, IE plural suffix *-es seems to correspond to the Uralic plural suffix *-t. Likewise, the IE 2sg verbal suffix *-s corresponds to the Uralic 2sg verbal suffix *-t. To these I would add the following:

    1. IE s-stem ending *-(V)s corresponds to the Uralic nominalizing suffix *-ta.
    2. IE *swéks "six" < **swékt; cf. Greek hektós "sixth" which (it seems to me) must come from IE *sw(e)któs -- otherwise the Greek form should be *hestós.
    2. IE dual suffix *-h1 corresponds to the Uralic dual suffix *-k(V).
    3. IE feminine sg. and neuter pl. suffix *-h2 corresponds to the Uralic diminutive ending *-ka.

    "5. "IE phonology had syllabification rules which favored open syllables" (again, why?)"

    I should have qualified this statement with "seems to have had". In any case, IE has many consonant clusters in word-initial position (i.e. complex onsets), which means that any such clusters in other positions would have a tendency to be treated as onsets as well.

    "If you're going to theorize, I except some degree of substantiation, not empty assertions."

    Fair enough. Hope this helps. :)

    ReplyDelete
  18. Rob, thanks for elaborating on issue #1. I accept.

    As for #2, what you seem to be suggesting is not that *h₁ simply has two equally distributed allophones (ie. my position), but rather that *h₁ has two distinct phonemic values based on no evidence whatsoever. Attention must be paid to the KISS principle: "Keep it simple, stupid."

    The PIE word for "sixth" is in fact commonly reconstructed as *sweḱstós and even that form may not even be the earliest form reconstructable for PIE. See Szemerenyi, Introduction to Indo-European Linguistics (1999), p.228.

    Rob: "IE feminine sg. and neuter pl. suffix *-h2 corresponds to the Uralic diminutive ending *-ka."

    I think that a comparison between PIE *-h₂ and the inanimate plural of Etruscan -(χ)va is a stronger match. One may then reconstruct Proto-Steppe collective ending *-hu (becoming Proto-Indo-Aegean *-hʷə), if anything.

    ReplyDelete
  19. Tropylium: "That's not the asymmetry I protest against, it's that there would be two different lenition processes going on in the exact same environment at the same timeframe.

    In my chronology there are 1500 years to play with during the Indo-Aegean period. So it's not necessarily the case that these things occurred simultaneously.

    ReplyDelete
  20. I said: "One may then reconstruct Proto-Steppe collective ending *-hu (becoming Proto-Indo-Aegean *-hʷə), if anything."

    Whoops, that should be *-hʷa. Minor detail.

    ReplyDelete
  21. Glen: "As for #2, what you seem to be suggesting is not that *h₁ simply has two equally distributed allophones (ie. my position), but rather that *h₁ has two distinct phonemic values based on no evidence whatsoever. Attention must be paid to the KISS principle: 'Keep it simple, stupid.'"

    A simple test would be to find instances of IE inflectional and/or derivational morphology where the (presumed) presence of *h1 in the affix does not cause lengthening of a preceding vowel.

    Like the vast majority of other languages, IE must have had roots/words which were considered to be vowel-initial, phonemically speaking.

    "The PIE word for "sixth" is in fact commonly reconstructed as *sweḱstós and even that form may not even be the earliest form reconstructable for PIE."

    I know that. My point was that I don't think how it's commonly reconstructed is correct. :)

    "I think that a comparison between PIE *-h₂ and the inanimate plural of Etruscan -(χ)va is a stronger match. One may then reconstruct Proto-Steppe collective ending *-hu (becoming Proto-Indo-Aegean *-hʷə), if anything."

    I don't know... it seems like the Etruscan inanimate plural is underlyingly /k_hwa/ or /k_w_ha/. I can't see any way that the IE ending *-h2 can derive from a labiovelar or a sequence of velar plus labial.

    ReplyDelete
  22. Rob: "A simple test would be to find instances of IE inflectional and/or derivational morphology where the (presumed) presence of *h1 in the affix does not cause lengthening of a preceding vowel."

    The onus of supplying this evidence rests on you. The optimal solution is that *h₁ is a single phoneme, even if showing allophonic variation, until proof of something more complex is successfully provided.

    Rob: "Like the vast majority of other languages, IE must have had roots/words which were considered to be vowel-initial, phonemically speaking."

    Assumption. No language necessarily requires "vowel-initial words phonemically speaking". Proto-Semitic also lacked this feature and considering the potential of intense language contact between Proto-Semitic and PIE during the Neolithic, I find that this interesting coincidence doesn't seem so coincidental anymore.

    PIE in fact does not have vowel-initial words, as shown by alternations like *h₁ésmi "I am" and *h₁smés "we are", or *h₁édti "she eats" and *h₁donts "tooth". These alternations only make sense when the word-initial glottal stop is accepted as a real phoneme, as shown by the reflexes in IE languages.

    Rob: "I know that. My point was that I don't think how it's commonly reconstructed is correct. :)"

    We plainly see from Latin sextus, Venetic sextio, Gothic saíhsta and Sanskrit saṣṭhá- that your view is incorrect and that *-s- is well attested in the PIE protoform.

    Rob: "I can't see any way that the IE ending *-h2 can derive from a labiovelar or a sequence of velar plus labial."

    Indo-Aegean *-hʷa would have become both Aegean *-ho (> Etrusco-Cypriot *-xwa/*-wa) and Old IE *-hʷa (> Mid IE *-hʷə > Early Late IE *-χ). Word-final labialization was neutralized after Syncope, hence there were no instances in PIE, nor could there be, of suffixes like **-gʷ or **-hʷ (**-h₃). Please also remember that I believe that the collective suffix *-h₂, like the nominative singular *-s and pronominal inanimate *-d succumbed to the Syncope rule despite the Suffix Resistance exception that affected monosyllabic affixes of the form -CV- by default.

    ReplyDelete