Re: [tied] The phonetic value of PIE *h3 and the 'drink' root.

From: Jens Elmegaard Rasmussen
Message: 14049
Date: 2002-07-17

Dear Piotr,

I cut a segment from your mail for subsequent comment:

> I propose to reconstruct the PIE
> underlier of the "-o:-/-i:-" ablaut as *-eiH-, where *H = *h2 or *h3,
> and posit the following phonetic developments:
> *-eiH-C > *-i:-C
> *-oiH-C > *-o:-C
> *-oiH-V > *-oj-V
> *-eiH-V > *-ej-V
> *-iH-C > *-i:-C
> *-iH-V > *-i-V
> For particulars see my article in IF 103 [1998], where I suggest a
> number of other examples of pre-laryngeally smoothed diphthongs.

I do not believe any of this. I have now read your article, and my verdict
is that all of the indications you interpret to support your rules can
easily have a different background, indeed, on balance, must have that.
The analysis that intrigued me the most was the rule *-eyH2 > *-i:(H2)
designed to give a nice underlying regularity for the weak form *-y-eH2-
of the devi: type. However, precisely for this type, the analysis is at
variance with the insight gained recently about the thematic vowel which
must be at play here. Adjectives of belonging are formed with "*-o-", the
old unaccented phonetic variant of which is -i-, or, as I say, even
earlier asyllabic -y-. Added to a consonant stem, this formed a fem. in
*-C-y-H2, i.e. a stem ending in three consonants. Since stemns ending in
three consonants (with a few special exceptions) are not permitted in IE
morphophomenics, a vowel was inserted, making the underlying sequence
*-C-yeH2-. Now in strong forms, where the accent was on the preceding
part, this was reduced to *-C-iH2, while in weak forms the added syllabic
case ending pulled the accent to the next vowel, i.e. to the
newly-inserted -e-, to give e.g. gen. *-C-yéH2-s. Note that the vrki:s
type is explained along with this, viz. by accent shift to express animacy
(therefore also the nom. *-s). the shift occurred before the reduction
product of "*-o-" lost its syllabicity, for it was in *-'C-i- that the
accent shifted onto the vowel to give *-C-í-s "an animate one belonging to
or resembling (the derivational base word)". Due to the initial accent
rule the accent could go no further to the right in this type.
You left it out here, but it is a major argument in your article that
we miss an ablaut gradation in a lexeme like *kruH2- 'blood'. Therefore
you derive the strong forms from an older *krewH2- which allegedly
developed into *kruH2- and so coincided with the weak alternant. Apart
from triggering the nagging question what then atteste *krew&-C- is, it
does not obliterate the very same "difficulty" for such stable root-noun
stems as *wid-, *dik^-, *dr.k^- which *must* reflect levelling in favour
of the weak-grade alternant. That causes me to ask, If the morphology of
*kruH2-/*kruH2- ain't broke, why fix it?
The long-diphthomg roots have a *very* strong tendency to turn up with
long vowel + y before a following vowel, a tendency not shown by root for
which the structure *CeyH- is known. Thus, roots of the structures *CeHy-
and *CeyH- behave in very different ways. There is, however, a certain
amount of common ground: First, the nasal present type made from both is
*Ci-ne-H- with the /y/ before and the /H/ after the infix; I take this to
indicate that the metathesis by which the originally suffixal -n- worked
its way into the interior of the root took a course that neutralized the
difference of the two root types here considered. Second, there is
metathesis of -VHy-CV- to -VyH-CV-, and, ancillary to that, later
derivatives may be based on the metathesized form (perhaps Sl. doiti and
poiti belong here? Lith. dial. sejù 'bind', Latv. kreju 'separate' cream
from milk, what else?) and Ved. dháyati are certainly best analyzed this
way, although it took me some time to come to that conclusion. I have
treated the matter at length in the first chapter of my book on IE
morphophonemics of Innsbruck 1989. I do not find that evidence addressed
in your paper.
Phonetically, I find it improbable in the extreme that /H1/ would be
retained where /H2/ and /H3/ are lost. Anatolian tells us that /H1/ was
the evasive one. But here, if the evidence is there, I accept anything; I
just don't see that evidence.
In the quote below, I fail to see the lesson to be learnt from bló:sko:
which is generally posited as *ml.H3- with an aorist *mel&3-/*ml.H3-
yielding /mole-/ just like eporon, eboron, estoresa, ekoresa, etc., the
funny "coloration metathesis" being a specific effect of /H3/ and that
alone. May I repeat that in my analysis IE *póH3-tlo-m and *póH3-mn are
perfectly regular forms, both having lost the root-final /y/ before two
consonants; for the difference between *-mn and *-mVn- note the homonymous
Gk. po:^ma 'lid, protection' as opposed to poimé:n 'shepherd'.


> reduplicated present stem *pib-e- would have come from *pí-piH-e-,
> with the high vowel lost in the reduced form of the "enclitic" root
> after a reduplication syllable (> *pi-pH-e-). I have so far been
> agnostic as to whether the phonation change was conditioned by the
> laryngeal, but I must say that the voicing effect of the Hoffmann
> suffix is a very attractive piece of supportive evidence. At any rate,
> under my analysis the thematic aorist *pi-é- comes from *piH-é-
> (presumably *pih3-e-), in which the voicing (if caused by *h3)
> naturally cannot apply. The unattestation of *peiH- is explicable if
> 'drink' was a root with persistent *o vocalism in its strong forms
> (like, say, *molh2-, cf. Gk. bló:sko:, émolon), not due to colouring
> by *h3. Hence also derivatives like Lat. po:culum < *poih3-tlom, Gk.
> po:ma < *poih3-mn, etc.
> Piotr
> ----- Original Message -----
> From: Jens Elmegaard Rasmussen
> To:
> Sent: Monday, July 08, 2002 1:18 AM
> Subject: Re: [tied] The phonetic value of PIE *h3 and the 'drink' root.
> Dear Sergei and List,
> let me comment on your stimulating mail:
> On Sun, 7 Jul 2002, sergejus_tarasovas wrote:
> >
> > [...] I would only like to note that the
> > analogical levelling proposed for Slavic seems a bit tricky to me,
> > especially considering deverbals like *pivo 'drink', *pijanU 'drunk'
> > and *pivIca 'drunkard', which don't look like recent formations.
> Then what *do* they look like? And if they are very old, how can we know
> they have not been refashioned on the way? To me they look like
> indications that there was a nucleus of truth to Martinet's notion of a
> change H3 > w under unknown conditions in PIE. Of course, we'd need to
> know what that nucleus was.
> >
> > Baltic (especially Old Prussian) indeed shows a strong support for
> > the *po:-, whatever be it's origin (though the accent is not very
> > clear):
> >
> > OPruss. inf. (<sup.) _p(o)u:ton_, inf. _poutwei_, _pou:t_ 'to drink',
> > 2pl imp. _poieiti_, _pogeitty_, _puieyti_, _puietti_ (lege *puieyti),
> > 2sg imp. _pogeis_.
> > On this base Maz^iulis reconstructs sup. *po:tun, inf. *po:t(wei)
> > (that's easy!) and 2pl imp. *po:jaiti, 2sg. imp. *po:jais resp. pres.
> > *po:ja- (he considers the imperative forms to be barytones, hence
> > unstressed *o: > unstressed *u: > (open) *u).
> You mean "non-barytones", as M. writes. That is contrary to Hirt's Law,
> but still may be correct, if analogical, reflecting as it does a
> productive stem-formation of younger reshaping.
> > OPruss. _poadamynan_ 'süsse Milch', if from *po:dam-in-a- 'drinkable'
> > (with dialectal merger of *o: and *a: like in _da:t_ 'to give'),
> > would point to an alternative West Baltic pres. *po:da-, probably
> > from *po:- + *-da-, cf. Lith. _vérda_ 'is boiling'.
> >
> > It's quite possible, that Lith. _puota`_ (acc. _puo~ta,_) 'feast'
> > also belongs here, if from deverb. adj. *po:ta: 'what is drunk' (with
> > acute->circumflex metatony, sometimes accompanying derivation adj. ->
> > subst.).
> Exactly! That's what nouns derived from adjectives by change of accent
> placing get if they are formed in so late a period that there were no more
> acutes to be handed out. It's all in the timing - that's why it hit the
> Slavic loanwords also (kny~gaN and all that).
> > As to the phonetic status of *h3, I'm still not sure that this _only_
> > example envolving *peh3i- is enough to state it was voiced.
> It's not based on this alone anymore. Most scholars however accept only
> this item, or at most Hamp's brilliant analyis of Celtic *abon- 'river'
> along with it, this being *H2ap-H3Vn- with the "Hoffmann suffix" of
> belonging (actually a compositional part forming mass nouns, including
> mass possession when they are bahuvrihis). By the change *-pH3- > -b- the
> Celtic loss of *-p- (Skt. gen. ap-ás, á:p-as) is prevented.
> Birgit Olsen has found a rather substantial number of examples of
> derivatives with this suffix which, when added to stems in -t- (or perhaps
> better, in -s-/-t-), or to stems in -k- (in part from hardened
> laryngeals), turn these into derivatives in *-don- and *-gon-
> respectively; thus Latin de-adjectival abstracts in -tu:s, -tu:t-is have
> synonyms in -tu:do:, -tu:dinis; and vora:x, -a:cis forms vora:go:,
> -a:ginis. Particulars are on their way in the press, you'll have to wait a
> litttle while. The new examples (some of which were in fact presented in
> the Pedersen memorial volume of 1994) have extracted píbati and afon from
> their isolation.
> Jens