Re: [tied] Re: IE lexical accent

From: enlil@...
Message: 33546
Date: 2004-07-16

Jens:
> I do'nt "have to recognize" any single one of them. Actually 2. and
> 3. are internal conflict. 4. makes no sense at all,

First of all, it's "don't". Second of all, I've already explained how
the fact that 3ps *-t is found medially while inanimate *-d never is
is the deciding factor for which stop gets voiced. Those in exclusively
final position will be voiced sometime in the Late IE period, as was the
nominative *-s [-z], something we both in fact agree to. As always, my
theory not only explains the voicing of *-s, it also is able to keep the
only logical etymology of that morpheme intact.

But... My theory just evolved again. Don't worry, just minor changes
involving MIE and eLIE phonotactics and vowel systems. Here is the
latest:


A NEW UNDERSTANDING OF SYNCOPE RESISTANCE
-----------------------------------------
After more analytical restructurings of my existing theory, I just
realized that all "clipped" suffixes (those that became asyllabic) are
all _word-final_. Nominative *-s, 3ps *-t, collective *-x and inanimate
*-d all qualify as word-final morphemes. That is, other suffixes are
not suffixed to them at the time of Syncope. (This would just imply that
the indicative *-i developped just after this in early Late IE, which
isn't much of a problem considering that I've always known that the
suffix developped around this time period based on other considerations
and logical chronology of events.) However, suffixes like *-no- are not
word-final (except for in the later vocative case) and are followed by
case suffixes and plural/collective markers.

So this means that "clipping" isn't an exception to an exception at
all. It's not even an exception. It definitely obeys Syncope. Now,
only the medial morphemes like *-no- seem to evade it. Since I've noticed
other stems, even without suffixes, keeping final vowels because of
phonotactic contraints after Syncope (to avoid triconsantal clustering
in final position), there are less "exceptions" by this and the small
residue that's now left can be explained by the already existent
restraints on permissible syllable shapes in eLIE.

That means that Syncope Resistance was the foremost strategy to stopping
dysphonic clusters, particularly in final position. The other strategy
was a-Epenthesis which _added_ the neutral vowel *a to break up clusters
that would otherwise occur in initial position such as *a-ktwa 'eight'
(> *okto:u).

What about the aorist *-s-? Well, it lengthens the preceding vowel like
the nominative does. This is because of the loss of a vowel by Syncope
on a monosyllabic sigmatic suffix and is nothing more than compensatory
lengthening. Now, if the aorist is truly a medial morpheme, why does
the lengthening suggest clipping as if it lost a vowel by Syncope?

Well, the paucity of Anatolian sigmatic aorists suggests that this
wasn't fully figured out yet. This'd be because it was only created in
a very Late IE, just a verb stem derived from an *s-stative noun. Before
this, one might presume that aorists were simply derived from duratives
by conforming the verb to the pattern of *CeC-a- in MIE (ie: the root
aorist). Tyrrhenian uses a similar verb ending to the aorist, as in
Etruscan's 'gerundive' /ar-as/ < /ar/.


NEW THOUGHTS ON *o/*e alternation in *po:ds/*pedos
--------------------------------------------------
Basically the new idea is that all instances of pretonic *a directly
become tensed to *e in early Late IE. This includes vowels that later
show up as *i as those in *i-stems and *i-reduplicatives. In this
new view, these *i's started out as eLIE *e. Any open-syllable *e's
were only _later_ raised to *i, perhaps in Mid IE.

By the way, the *e in eLIE *pedás was in a _closed_ syllable because
it is part of a CVC stem. The word is properly syllabified not as
**/pe.dás/ but as */ped.ás/. In the eLIE perfect, 3ps *bHebHára is
also properly syllabified as */bHebH.bHár.a/ while in the would-be
*i-reduplicatives such as *bHebHérati it was */bHe.bHér.a.ti/. So
naturally, the *e in an open syllable will be phonetically longer and
more prone to rising to *i.


Jens:
> We need an argument for the choice that the presumed reductions of *so
> to *-s and of *to to either *-d or *-t postdate the assignment of the
> accent.

Not only does the contrast of *-s and *-d mirror both phonetically and
semantically the demonstratives *so and *to-, and not only is the
suffixing of former demonstratives attested in other languages (even
Scandinavian ones!), but also the very nature of QAR itself suggests
overwhemingly that the loss of final vowel in these endings long
postdated the assigment of accent. Quasipenultimate accentuation took
shape in Mid IE.

Luckily for me, there are yummy pecularities in accent placement. While
it is largely predictable as to whether the accent is found on penultimate
or antepenultimate in MIE via QAR, there are some minor cases where the
accent had to actually be learned. Egad! Namely, in paradigms of animate
*i- and *u-stems. If we have nominative *héwai-sa 'bird' (> *xewis), then
genitive *hawéi-sa cannot be predicted except by understanding that the
genitive attracts accent one syllable ahead. For consonant-ending stems,
the genitive's accent is entirely automatic because the attraction of
accent is accompanied by a vowel before the suffix -- *wát:an 'water'
(*wodr) and *wat:an-ása (*wednós ~ *udnós) are perfectly predictable.

What's going on? Well, we need *hawéisa to get later *xweis. It appears
that *i- and *u-stems get the genitive *-sa without intermediate *a just
like vowel-ending stems for the simple fact that *i and *u are vocalic
glides in their own right. The *-a- is a way of breaking up the stem from
the suffix ending but in *i- and *u-stems it was unneeded at the time
that these case suffixes had been created in the first place.

So... why is there a difference between the genitive which has this
interloping *a and the nominative which doesn't? Why are their accents
different if they look the same? The simple reason is that the genitive
was created BEFORE the nominative in IndoTyrrhenian. As in Uralic, the
nominative was originally unmarked for both genders while the ancient
accusative *-m was restricted to animate usage. After cases were
synthesized from pre-existing postparticles using the euphonic go-between
*a for consonant-ending stems (eg: Steppe *si "out of" > *-(a)sa),
IndoTyrrhenian decided to mark animacy for _both_ strong cases, rather
than just the accusative, by postposing the general demonstrative *sa
after the noun. At this stage, it was not a true case "suffix" however.
It was as yet only a particle placed after the bare animate nominative.

When exactly speakers considered it a true _suffix_ is hard to tell but
it definitely was not a suffix until Mid IE when QAR had developped
because the accent pattern evidently recognizes that the nominative is
different from the genitive as in the *i- and *u-stems. This would not
be possible if nominative *-sa was thought of as just another case
ending.

So what happened exactly? Simple. An initial light accent in the earliest
stage of Old IE gave way to a strictly penultimate accent some time around
early Mid IE. Once postposed *sa and other particles were treated as
suffixes, it distorted the penultimate accent since these morphemes were
not originally part of the word. The language then came to allow
antepenultimate accent in addition to penultimate as a result. Thus was
born the Quasi-Penultimate Accent Rule (QAR).

The inanimate suffix *-d and 3ps *-t(i) also have the same source and
explanation. They too were from the same postparticle *ta acquired in Mid
IE that ended up marking first 3ps duratives and then inanim. pronominals,
in that chronological order. It's without surprise that they also show
the same antepenultimate accentuation whenever they are used.

In IE itself, the lingering evidence of what I'm saying is suffixes that
'fail to attract accent'. A desinence that fails to attract accent implies
a desinence that postdates QAR because it is invariably a particle. It's
brilliant really because it explains all those pesky asyllabic suffixes.
Even mediopassive *-r... again an MIE particle *hWar (a bare stative
literally meaning 'done' but used to mean 'via, by which'). How about
indicative *-i? Yep, originally MIE's endingless locative *ei meaning "at
the time, here".

The rest is... erh... prehistory.



> That is not what I see when I check on that.

What you see concerning EA and Uralic is not what most others see
either.


> We should indeed have that, why not? The stem is *H1es-, and it has
> a lexically given accent on its final vowel, i.e. *H1és-. The
> addition of a syllabic ending then pulls the accent to the following
> vowel in the usual manner: **H1es-ént > *H1s-ént, indicative *H1s-
> ént-i.

Ah. I was confused by your use of stem to include the suffix in this
case. However, we have to recognize the stem as we do in IE itself.
So, in all cases of _paradigmatic alternations_ (please note the
emphasis) a stem ALWAYS contains a syllable. We just never see anything
like **pdos but we can certainly observe *pd when it is in a derivation
to form a _new_ stem with its own paradigm. Even still, in that derived
paradigm, we again never ever ever see the stem reduced to asyllabicity.
It STILL has at least one syllable.


On Syncope:
> Many languages have both. Your irate outburst is like saying
> (better, shouting) that the last vowel of Old Irish tarathar 'auger'
> cannot be anaptyctic if there is syncope of the second vowel in
> dat.pl. tarthraib; still Welsh which has none of this has taradr.

Alright, but in this case Occam's Razor forbids me to consider anything
but the simplest solution until some motivation based on the facts urges
me to adopt a more complex one. What shows that IE must have both
Syncope and clustering within a syllable? All I can see and all that
ever appears to be needed is a simple CVC structure for MIE syllables.

So, don't be confused. I don't reject your Celtic examples as impossible
here but I just don't see the need for it in IE when everything can be
explained without this further complexity. I mean, how could one tell
whether one initial cluster is truly a cluster and which one isn't. It
would be impossible. So what 'careful analysis' shows anything? What
is shown by any analysis is that Syncope did exist and that some
clusters are indeed the product of Syncope. I admit to generalizing
Syncope for all clusters but... that's Occam's Razor in practice.

What we gain from this generalization of Syncope to ALL clusters is
a nice CVC syllable structure for MIE for ALL syllables that works
just fine. Further... a-Epenthesis and Suffix Resistance would appear
to be two strong testimonials to a _constraint_ on clustering! So I
simply can't follow your point of view because it loses too much in
exchange for outright ignorance and confusion.


> Were there no simple consonantal endings in the morphology?

In a sense, yes. However phonotactics forbade the direct application
of a suffix right after a consonant. For example, the MIE durative
requires the 'thematic' *e and the aorist a thematic *a in between the
singular endings and the verb stem even though the suffixes are
technically just consonants (Eg: *és-a-m 'I am'). The thematic vowel
is not only signalling aspect, it is also there for more practical
reasons. There is also the very ancient suffix *-m in the accusative and
it too requires *-a- before it in consonant-ending stems as do most
other case endings that aren't consonant-only like the genitive *-(á)sa.

Me:
> They must surely have had vowels before and so what then was the rule
> _at THAT stage_??

Jens:
> We cannot know.

Ach, _you_ cannot know because you can't be reasoned with. Every time
I run you into a logical brick wall, you say "We cannot know" as
predictably as a rooster crows in the morning.


I fortold:
> (Prediction: Jens will say "we cannot know", whereupon I will think
> "Load of horse doodoo").

Jens:
> You *have* learned something by now. I still have to learn to fathom
> the depth of some of your more subtle arguments.

Load of horse doodoo.


> The language itself tells me that there is accent movement with
> morphemes which have allomorphs like *-so, *-ey, *-eH1, *-oom, [...]

And this observation doesn't demonstrate a clear conclusion, only a
meaningless pattern (if it really is a pattern considering that the
nominative has a zero-allomorph in *kwo:n and that the perfect endings
are all rejected under your ad hoc analysis). All it does is encourage
anyone to imaginatively find their own differing solutions from the
same "pattern" because there is no inevitable conclusion that can come
from any of this.

You always take from examples specifically designed to be unencumbered
by clarity so that your solutions can be unencumbered by reality.


> Then there is no basis for a vowel, and no basis to make a rule from.
> I only wish you had added an argument.

IE 1ps *gHwenm likely comes from MIE *gwénam and causes no problems
because I now see that athematic verbs can have once been thematic
by way of Syncope.


> And what might he say? Miguel's Tour of Proto-Nostratic Grammar
> (hoping for your permission, Miguel) says this

Whatever.

At any rate, in regards to the pronominal endings in my correlative
ProtoSteppe and Boreal stages, we can indeed say that these endings are
"consonant-only" as long as we grasp that go-between vowels are
absolutely unavoidable by way of the limited allowable syllable shapes
of those protolanguages. I don't agree that Eskimo-Aleut just strung
a whole bunch of consonants together in Kartvelian-style without vowels
in-between and these analyses don't even require us to do so.


> The Tour also reconstructs a plural in *-t for the noun.

That's nice but it's just wrong. It has to be *-it because it explains
the palatalization in Altaic of resulting *-r^, it shows up in Tyrrhenian
as *-er with *e (Yes, we SEE that vowel in /clen-ar/ 'sons').

This is all just a fantasy of yours, Jens, to hold onto a corrupt
analysis of *-es. My theory doesn't have problems with the form. It
doesn't need to come up with a smokescreen like "the *e just popped out
of thin air somehow". Of course, we have no choice to explain *-e- here
but my theory just analogically dissimilates a vowel that would have
otherwise have become *o in IE. There are even other cases of an
expected unaccented *a becoming *e in eLIE, as in *pat:ása > eLIE *pedás
> *pedos.


> The same ending is asyllabic in Eskimo-Aleut (*-t, morphophonemically
> a dental spirant with subphonemic voicing, i.e., "edh").

No. It's still in reality *-&t. The schwa is what remains of Steppe *i.
But this is a matter of nitpicky analysis. Grammatically it may indeed
be said that it is a single-consonant and *& is just a 'filler'. I just
don't want it to be said that there is no vowel in-between.

Ultimately however, *i is part of the morpheme in Proto-Steppe because
it is *i that is reflected in the resultant forms, not *u and not *a.


= gLeN