From: Jens Elmegaard Rasmussen
Message: 21245
Date: 2003-04-24
On Thu, 24 Apr 2003, Glen Gordon wrote:
>
> Jens:
> >Well, fine up to a point. The basic argument against accepting the
> thematic vowel as "old" (meaning as old as the ablaut or older) is
> the idea that they would have been lost if they were that old. That
> makes sense if (1) there is a rule deleting all unaccented
> >short vowels, (2) nothing else develops into the thematic vowel, and
> (3) the thematic vowel is always unaccented. Prerequisite (2) is
> unknown, and (3) is false. I'll come back to that.
> Glen:
> The "thematic vowel" that I refer to involves ablaut alternations in
> *e and *o. This phenomenon is the tell-tale sign of the earlier schwa
> *& and is most usually unaccented. So accented "thematic
> vowels" don't really count since they are mostly from eLIE genitival
> constructs that never had contained *& because they never lost the
> original accent on the ultimate syllable. (3) is not as false as you
> think it is, (2) is known, and (1) is true except in some
> special cases: Paradigmatic Resistance and Suffix Resistance.
Jens:
I am fully willing to restrict the debate to the interchange of -e- and
-o- in stem-final position. I guess that descriptive definition of the
thematic vowel is one we can all agree upon. I do not accept that it can
be known that there was only one source of the thematic vowel. On the
other hand, I have no knowledge of any indication that it had a variety
of sources, but we cannot make conclusions about our investigation
before we carry it out.
Glen:
> The schwa *& which is behind the thematic vowel surfaced in
> genitival derivatives with an accent shifted to the initial,
> in unstressed particles, in verbal suffixes and vowel-final
> monosyllabic suffixes that carried on MIE *& (Suffix Resistance), and
> in "animatized" suffixes (eg: the animate collective *-&x from
> inanimate *-x). It is a seperate vowel from the others of its time:
> *i, *u, *e and *a.
Jens:
As accept the last sentence: The thematic vowel differed from other
vowels at some point in time, for the others did not behave this way. We
are closer to being in agreement than perhaps either of us would like. I
understand of course the scenario whereby the vowel of genitives in *-os
could be reinterpreted as nominatives in *-o-s, this offering a
stem-final vowel to be used in adjectives of much the same meaning as
the genitive itself. But I encounter very grave difficulties when I have
to visualize this really working, for there is so much against it that I
have ended up scuttling the whole idea: If the thematic vowel were in
fact the same vowel as that of the genitive singular, I would expect the
vowel of the genitive to act in the same way the thematic vowel does,
and that is patently not the case. The vowel of the gen.sg. is not
ablaut-resistant, but goes to *-s when unaccented. Also, when the
genitival sibilant meets the thematic vowel, the latter is -e-, not -o-;
that is seen from forms like *ésyo, *kWésyo. That the substantive has
*-osyo is not really evidence, for the old interplay -e-/-o- is hardly
retained at all in substantives which have generalized -o- except for the
vocative *-e and the fem./coll. forms where the e-variant was coloured to
yield *-aH2(-); the pronouns, on the other hand, display much of the same
allomorphy as we see in the verb, and apparently by the same rule. I also
have difficulty imagining just how a genitive morpheme could end up in a
verbal stem. And even if that could be imagined, I find it difficult to
understand how one can be sure about it.
Glen:
> It has little to do with Mid IE's unstressed vowels since these
> were mostly zeroed by Late IE. Schwa did NOT arise in cases of
> Paradigmatic Resistance (eg: *p&t:as& > *pet:as, not *p&t:as which
> would otherwise exhibit lengthening before a later *d, producing
> **podos). The schwa only surfaced in special environments and could
> not be zerograded further. However it could be fullgraded to *e in
> paradigm alternations.
Most of this is given in negative form, so we are in basic agreement here
too, for certainly I do not defend any of this in exactly the form you
give it. Your wording suggests that you operate with "deep-structure"
thematic vowels which do not always surface. I feel that would need some
further backing. The root vowel of "foot" is not a thematic vowel in my
world. I simply do not observe a voice-governed alternation e/o in root
segments. If there was ever a vowel following the sibilant of the gen.sg.,
I fail to see the basis for it, nor its relationship with the thematic
vowel. For the time being I must consider these preforms unmotivated.
Glen:
> Evidently, thematic vowels as I define them above appear to have
> "survived" ablaut, if we go by your scenario. However, by granting
> them ancient status, we obliterate the regular penultimate accent that
> explains the later mobile accent and we complicate tenfold
> the processes under which Mid IE had lost unstressed vowels,
> among others. The loss of unstressed vowel becomes a suspiciously
> selective rule. We also end up with a much looser syllable
> structure, allowing for two consecutive vowels, which surfaces in your
> thought experiments below.
Jens:
Indeed I believe they survived ablaut unimpaired. The loss of unstressed
vowels skips stem-finals. They must do that for a reason, but that has not
been found. If it is an effect of stem-final vowels per se, it must
reflect something more which we cannot see; that is not alarming in
itself, for most things must have changed and much will have been lost. It
would work if the vowel-less endings of strong paradigm forms were once
words containing vowels which have been lost; that would mean adding an
earlier round of ablaut loss before the generally observable é/zero
ablaut. That could well be the case: if *gWhén-m is really "a killer (am)
I", one would like an older form of that to have had a _word_ meaning 'I'
as the second part; that would in fact demand a word boundary between the
two elements and would thus motivate the analysis independently. Then,
whatever material developed into the thematic-vowel alternation could have
retained its voice-governed allomorphy. It does not arrive at any obvious
elucidation, but perhaps that just can't be helped.
Jens again:
There is a different possibility which none of us has apparently really
tested yet, although I have been driven in that direction by thinking
about vrddhi derivatives and subjunctives, but only to stowe it away as
taking me too far afield: What if the thematic vowel is not the final part
of the first of two words in close juncture, but instead the first part of
the second of two such words? Then, the derivative *kWr.tó-s 'made',
derived by thematicization from the agent noun *kWér-t (in compositional
final -kr.t- in Skt.), would indeed be (pre-ablaut) *kWer-t-ó- meaning
'what a maker has' (very close to your interpretation as an old genitive);
but, instead of regarding this as a stem-elaboration, one could take it to
be two original words: *kWer-t + *o-z meaning "a maker - his", if the stem
of the second word is taken to be a third person possessive, and the
syntax is taken to be a Gruppenflexion-like collocation of an uninflected
agent noun followed by a possessive, the entire combination being then
inflected in the nom.sg., for it is "what the maker has" that is now the
subject of the sentence (or an apposition agreeing with it). In THAT case
it would be far less strange that the thematic vowel acts in a special
way, for then it is an earlier word-initial vowel, and THAT would bring it
in line with common experience that initial syllables are often better
preserved than others. Irrespective of the accent we can observe, the
limits of variability of the IE vocalism strongly suggest that flexives
once stood in an added word of their own. For in the root one can have any
vocalism, long or short (though /e/ is massively hyperfrequent); in
suffixes we find only short /-e-/ (or what comes from it) or no vowel at
all; in desinences, now, we again find a variety of vowels, but only short
(gen. *-os, 3pl *-ent, instr. *-bhis, loc.pl. *-su). That means the
desinential segments behave loke roots in the sense that they do not show
total monotony in their vocalism, but still like reduced entities by being
so short and by never having a long underlying vowel. That points to the
inference that the IE inflected word is basically a juxtaposition of two
words of which the first could take modifying suffixes. The thematic vowel
is located in the middle and has a weight problem. Now, we just have no
other way of knowing how much weight a vowel located in that slot should
carry except by looking at it. It shows a status of gradation which looks
like a middle thing, and it stands in the middle. I think we should accept
that as an objective fact. If it does not fit our expectations then
perhaps they ought to be revised.
Jens:
> >There is very solid evidence that the thematic vowel meant something,
> [...] It formed ordinals from cardinals, and subjunctives from
> injunctives, both without any apparent assistance.
Glen:
> This isn't solid evidence. Its employment in derivation is so wide and
> varied, as by your admission above, that there can be no
> reasonable meaning attributed to it, nor has there. It's weak.
Jens:
Ordinals are adjectives made from cardinals, i.e. modifying numerals.
Subjunctives would then be, I suggest, modifying verbal forms, i.e.
originally dependent-clause forms marked for what they do, namely modify
main-clause verbs. That would explain the use of the very same morpheme
for both functions which can thus be reduced to one. Many American Indian
languages operate that way, I am told; I only know it from Eskimo myself.
Other suffixes that end in a vowel also exhibit the thematic-vowel
alternation, but there is no reason to believe that they all contain the
same element. But if they just represent something standing first in the
second part of the two-word collocation, that may give them the same right
to behave that way. Then, just as you would not demand a communality of
meaning among all words that have changed IE p into Germanic f, there
would be no call for a common function in everything containing the
thematic vowel; but ther _is_ an apparent agreement in function between
suffixes that consist _only_ of the thematic vowel.
Jens:
> >That's what we find in adjectives in *-ró- and participles in *-tó-
> and in present stems in *-sk^é/ó-. Accent on this vowel is here
> feeding the ablaut.
Glen:
> Yes, the accent _is_ feeding the ablaut, because adjectives like those
> in *-ro- and *-to- are descendants of genitival derivatives in eLIE
> *-as. They have nothing to do with thematic vowels because they never
> contained *& at any time in their existence.
Jens:
Well, I'd say, by the only sensible definition of the thematic vowel I can
figure out they do indeed contain just that.
Glen:
> Verb stems in *-ske- are late and only mimick other verb formations
> where zerograding of the stem already had occurred. I'd say that an
> accented alternating thematic vowel is unusual. Another thing unusual
> about the affix is its -CCV- pattern which lacks a coda
> despite being accented. It seems reasonable to presume then that this
> is not an ancient suffix because it doesn't follow the rules we
> attribute to a more ancient layer of IE.
Jens:
Now, this is one of the rare occasions where masks are off. We find the
descriptive phenomenon generally termed thematic vowel accented and
unaccented, and we find it in suffixes of different structures. I know of
nothing that would disqualify half of the evidence, I only observe you
doing that. It would be very strange if verbs in *-sk^é-/-sk^ó- and
*-yé-/*-yó- were to be explained as analogical on a type in *-é-/-ó-
alone, for there is no old type of that structure. The tudáti type is an
innovation based on the aorist (reinterpreted as imperfect), and the
aorist type inj. vidát (ind. ávidat) is only secured for two verbs, both
times replacing the middle voice, and so obviously reflecting a
normalization of the expected 3sg mid. *wid-é, *H1ludh-é to *wid-é-t,
*H1ludh-é-t, now apparently of active voice (but of unchanged meaning).
And of course an aorist type would be a funny model for a present-forming
suffix anyway.
>
Jens:
> Concerning thematic alternation:
> >I don't know why you are so bent on calling it length, though.
Glen:
> Lengthening by voiced segments takes less pleading than rounding by
> voiced segments. Such lengthening is strongly backed by
> real-world examples. I don't see any evidence for rounding on the
> other hand and must at the very least be less common.
>
> Therefore "lengthening" (plus vowel shift) is a more logical
> theory. It works in coordination with the vowel shift *a > *o that I'm
> already aware of. Thus we may theorize that lengthened *&: first
> merged with *a before finally becoming *o while *& merged with *e (or
> *A > *a when uvularized).
Jens:
We are not all that much in disagreement here. I now believe Hirt was on
to something when he insisted that the change of e to o was the effect of
a musical accent. This can bring the two o's in under a common heading:
Since the IE accent had high tone _and_ stress, unaccented e-vowels had a
lower tone. Now, the vowel /o/ (of whatever exact brand) has a distinctly
lower tone than any kind of /e/ (it's the "second formant" I'm talking
about). Since also voicing lowers the tone, then the change of e to o
before a voiced segment as observed in the thematic vowel could be an
event of simple assimilation, for the lowered variant of /e/ would have
been brought closer to /o/. Calling it length is not far off the mark, the
proper parameter being rather volume.
>
Jens:
> >I'm struggling to follow ... Many, very many, thematic stems are not
> stressed on the initial. What is funny here? You funning us?
Glen:
> Yes, I'm aware. I ain't funnin' nobody. However, I suspect that we
> might be viewing "thematic stems" differently, judging by your
> comments on the stems *kWi- and *kWo-, both of which you call
> "thematic" (which is true in some ways, and not in others). In
> the quote you were responding to, I'm using "thematic" to mean
> only those stems ending in any other vowel other than semivowels *i
> and *u. If I understand my IE grammar well enough, that leaves only
> stems in accented or unaccented *-o-. Only unaccented *-o-
> applies to the topic of eLIE *&.
Jens:
I respect your restriction of the discussion to -e-/-o-. I protest against
a crude periodizing giving names to stages we are just going to find yet.
If we find 98 changes, there are 99 stages, and I would not know how to
asign names to them all, or to any given one of them.
Jens:
> >Ironically, if the thematic structure is subjected to normal
> >ablaut, it comes out the same: Pre-ablaut *perk^-sk^é-t, 3pl
> >*perk^-sk^e-ént would become first *prk^sk^é-t, *prk^sk^e-ént; then,
> with the initial accent rule (which I see is accepted by
> >Miguel) *prk^sk^é-t, *prk^sk^é-ent; then, with continued ablaut
> reduction, *prk^sk^é-t, *prk^sk^é-nt; whence finally, with the
> >thematic vowel rule, PIE *prk^sk^é-t, *prk^sk^ó-nt. The fact that the
> thematic vowel is never followed by the accent means that it must
> date back to a time preceding the initial accent rule.
Glen:
> You see, I object to the syllabics of the above (eg: *perkskeent ??).
> I really don't see allowance for two consecutive vowels in any stage
> of preIE. The further back in time I go, the more I see *CV(C)CV(C)-
> as the norm, the same norm used by Uralic and Altaic, if not also by
> proto-Tyrrhenian. It looks to me that *VV was only allowed in the
> latest form of IE (note Miguel and I on case endings with so-called
> "circumflex" accentuation). Your reconstruction is too indulgent.
Jens:
I am not saying there _never_ was a consonant between the thematic vowel
and apparently vowel-initial flexives, I am saying that we cannot see any
for the preiods we are talking about here, at least not on the sole
evidence of IE which is all we are talking about. I do not think the
words structure constraints of Uralic and IE are congruent, it looks more
as if they have each retained different parts of a larger whole. I'm
sorry, but I would be indiscrete about the work of colleagues if I express
myself more specifically on that matter. We agree on the circumflex of
course. I don't see who any of us is to forbid the language to present
certain structures before we go and look. Much progess, by the way, has
been accomplished by scholars who were the first to stop doing just that.
Glen:
> That's perhaps another reason to have thematic vowels postdate
> quantitative ablaut -- The rules for syllabic shape become much
> less liberal for armchair linguists. Yet another reason is that it
> makes no sense why your preAblaut *perk^-sk^e-ént shouldn't become
> *perk^-sk^-ént instead, after what I presume to be the stage
> representing late Mid IE and the loss of unstressed vowels. ??!
Jens:
Now, back to square one: I was talking about stem-final vowels and their
apparent resistance to ablaut reduction. In pre-ablaut *prek^-sk^e-ént (if
there was such a form, which I do not insist there was) it is the middle
-e- that is stem-final, so that's the one that is sacrosanct. Then, after
ablaut (part one), the form is *prk^-sk^e-ént; on that basis, the initial
accent rule would create *prk^-sk^é-ent, and the ablaut (part two) would
turn that into *prk^-sk^é-nt which, with the thematic vowel timbre rule
would produce the output *pr.k^-sk^ó-nt which everybody accepts. It's all
because of the special character of the subject we are talking about, the
thematic vowel. That is why we are talking about it.
>
Jens:
> >Now, ablaut worked also *after* the initial accent rule, because it
> was fed by it.
Glen:
> No, because accented thematic vowels as per your examples above do not
> derive from *& unless we lazily allow an accented *& as well (but
> then, on what firm basis?).
Jens:
Call me what you want, I sometimes am lazy. I do in fact allow an accented
thematic vowel. That is based on that half of the evidence you disregard
by an authority I do not accept.
Glen:
> More vowels, more freedom for syllabic shape, more possibilities, and
> before you know it, the language is all over the place because we
> weren't diligent enough to reconstruct it properly using the
> least amount of hypotheses.
>
Jens:
I am not making unnecessary assumptions, I perhaps just don't forget so much.
Jens (on a different cue):
> >Oh yeah? I must conclude the opposite.
Glen:
> Always the opposer. Congratulations: You are proof that world peace is
> not achievable >:)
Jens:
If it can produce world peace I surrender without a frown.
Jens:
> >Why would a genitive give up its -s and create a verbal stem out of
> the stump? What is a truncated genitive doing in *bhéro-nti 'they
> carry'?
Glen:
> No, no. First of all, the verbal thematic vowel derives from the same
> vowel *& as in the noun, but it originates from something a little
> different. The mLIE equivalent of *bHér-o-nt-i just after Acrostatic
> Regularization was *bHér&-nt. The thematic vowel here derives from the
> default 3ps *bHér&-t which before the add-on *-t was seen as
> "endingless" in eLIE *bHér& (*& survives by
> Paradigmatic Resistance, note 1pp bHerém&s, and by Suffix
> Resistance because it is a monosyllabic suffix). So in verbs
> with 3ps *&, the schwa spread by analogy to all persons to create the
> "thematic" verb.
Jens:
That is just a guess. I find no reason to believe it, let alone proof that
I have to. Apparently you are not satisfied with the language as you find
it, for you explain a totally different language and then postulates a
magical change of it into what you find. That would seem to leave the
rules vacuous.
Glen:
> Getting away from verbs, the eLIE genitival construct was used to
> convey an adjective or a descriptive noun. It was as a descriptive
> noun that the stem came to have initial accent, thereby ending in a
> reduced *-&s, while the correlating adjective retained accent on the
> final syllable *-as. The accent change was a simple way of
> distinguishing inflected noun from an originally uninflected
> adjective.
>
> At this point, the genitival constructs both nominal and adjectival
> were misanalysed as vowel-final stems plus nominative ending *-s,
> giving noun stems ending in *-&- and adjectives in *-a-. So "a
> genitive gives up its -s and creates a verbal stem out of the
> stump" because genitival constructs just looked like animate
> nominatives. The adjective became inflected after this and came to
> agree in case with the noun it modified using its "stumpy" remains.
Jens:
This is the part I find makes sense, only I do not accept it; I believed
just about the same thirty years ago, but I can't find a free spot for it
now, there are red lights all over.
>
Jens:
> >I see no acrostatic regularization in the thematic class (I suppose
> you mean structures like *bhére-ti and *wérg^o-m by that, but I do
> not feel sure about it, for it means disregarding *prk^-sk^é-ti,
> *gWm.-yé-ti, *wid-H1-yé-ti and the noun types *yug-ó-m, *k^m.tó-m,
> *H2ug-ró-s, *mr.-tó-s, even mid.ptc. *dhugh-m.H1nó-s, and can you
> really have overlooked all that?).
Glen:
> No, I didn't overlook it. Your list gives me a lot to talk about.
> Verbs in *-ske- and *-ye- were formed only in the latest stage of Late
> IE. They only follow ablaut patterns because they mimick other
> derivatives with zerograded verb stem that are ancient enough to have
> operated under the ablaut. Nouns like *yugóm, *xugrós and
> *mrtós are genitival constructs out of the eLIE case endings *-as and
> *-am, never containing *&. The numeral *kmtom is actually short for
> */komtx kmtom/ "tens of tens", and *xugrós & *mrtós are nouns of a
> late type derived directly from adjectival forms, hence their ultimate
> accentuation. This *dHugHm.hnós is again an adjectival
> form which again connects with the genitival constructs.
Jens:
You just can't know any of that. And there are no problems if accented
thematic vowels are simply accepted. You are divorcing the thematic types
from each other. And you are even reprimanding the language, refusing to
accept large parts of the evidence it offers.
>
Jens:
> >If Indo-European stem-formation is described for what it shows, it
> should be accepted that a stem could also end in a vowel.
> >[...] This of course begs the question: What is so special about
> stem-final position that a vowel positioned there will go haywire and
> make this kind of spectacle of itself? I do not know,[...]
Glen:
> It is the very reason that the stem-final vowel acts in different,
> otherwise inexplicable ways that shows us that we CANNOT accept
> that the thematic vowel predates the same quantitative ablaut it
> defies.
Jens:
That is not the only possible solution to thhat problem, and, as it turns
out, not the right one to select. If the thematic vowel does not behave
like other vowels, it can also have been just sufficiently different from
other vowels to be excepted from their rules. In fact there is ample room
for that solution.
Glen:
> I fail to understand how vaguely waving the thematic vowel to
> "the dark corners of prehistory" should give us the coherent
> understanding that I'm actively seeking via my own account.
> I don't see that coherent explanation coming from you.
Jens:
We cannot expect to find an answer to everything, there will *always* be
something whose motivation predates everything we can control by rules. We
are of course approaching a point of complementarity: We can only make
inferences about A, if we stay agnostic about B, and vice versa. In
practice, however, I have not seen such a dilemma yet. But I do not feel
obliged to deliver the full story of the thematic vowel on a silver
platter. I only feel obliged not to tell lies about it, and that's
difficult enough.
>
Jens:
> >It is the opposite that follows logically: Since even "the earliest
> rules we know" produce structures that were not hit by
> >the thematic vowel alternation, such structures missed the train -
> they are too young.
Glen:
> But this relies on an assumption that thematic vowels are ancient,
> without first showing that this is so. Given the big picture, it seems
> to make all the rules that much more complicated and defies Occam's
> Razor. So it can't be logical.
Jens:
I don't think a scenario of their not being ancient has been left
untested, and none has proved viable or even possible. Well, then they are
ancient. I don't think Occam himself could find fault with this.
>Jens:
> [After giving examples of the thematic vowel evolving into PIE:]
> >Now, accent has no part in it, and a three-quarter-genitive appearing
> as a verbal stem is somehow unappealing to me. And I wonder how
> pronouns can have *tóm, *tód, *tóy with the same rule
> >applying to monosyllables, if it is supposed to apply only to
> unaccented vowels.
Glen:
> One, the verbal stem has nothing to do with the genitive. The End.
>
Jens:
Right, it hasn't even got that basis.
Glen:
> Two, the pronouns are semi-enclitic, that is, they were typically
> unaccented in a sentence unless accent was necessary for emphasis or
> whatever other reasons I haven't thought of yet.
Jens:
So pronouns with accented thematic vowels are just ill-mannered, and
verbal stems accenting their thematic vowels are ill-mannered, but you
know better and can't be fooled, not by all the evidence in the world? I'm
sorry that is the impression it leaves me with.
Glen:
> So, the paradigm of *to- is largely built on unaccented forms with
> *t&- which derive from the accented forms in *ta-. Likewise, *kWo- is
> built on forms with earlier unaccented *kW&- (accented *kWe-, not to
> be confused with forms in *kWi- whose differences I shall explain some
> day if everybody is nice to me :P).
Jens:
I can wait.
Glen:
> In English, "this" is pronounced [DIs] in isolation but I might say
> [(D)&s] in a typical everyday sentence that I speak and [(D)iz]
> for "these". Probably a thousand years from now, linguists will
> wonder why Postmodern English case endings seem attached to
> unstressed forms rather than stressed forms. So anyways, back to IE,
> *tesyo is from unstressed *t&sy&: (which derives from the form
> containing stress, *ta-sy&:).
Jens:
Such things happen, sure. However, you don't know that it happened here,
and you don't need it if you follow me and get cracking at the whole
language and not only the selected form you invited to the party. I wonder
what your friend Occam would have thought of this.
Jens