Infixal /o/

From: elmeras2000
Message: 30940
Date: 2004-02-11

--- In cybalist@yahoogroups.com, "P&G" <petegray@...> wrote:
> > The initial *o- of a small number of secured formations is the
same
> > element as the medial *-o- of the numerous formations of the
types
> > *tomH1-á-H2 'cut', *tomH1-ó-s 'cutting, sharp', *mon-éye-
ti 'makes
> > think' (caus.), *tór-mo-s 'hole' (*terH1- 'bore'), *kor-mó-
> s 'stump'
> > (*ker- 'cut'). This element, which is generally treated as if it
> > were a special appearance of the root-vowel, behaves
> > morphophonemically as if it were a consonant.
>
> I 'm not attacking, just seeking clarification.
> In what way is it the same element as the -o- of the ablauting
vowel? Do
> you mean just phonetically the same? Or something more?
> How does it behave morphophonemically as if it were a consonant?
This
> implies that there is some morphophonemic pattern we can observe.
What?
>
> (Or should I wait till I can get my hands on that article?)

The -o- of the causative-iterative (*mon-éye-ti, *H2nok^-éye-ti) and
of nominal types like *bhor-ó-s, *bhor-á-H2, *pórn-a-H2 is NOT the
root-vowel in disguise. It is an element squeezed in next to (right
before) the root-vowel, after which the root-vowel itself has been
deleted (because the accent was on the suffix).

Real vowels in unaccented position are either lost (*pér-tu-s, *pr-
téw-s) or, if they were originally long, they attract the accent
(Ved. 3pl táks.ati from IE *ték^T-nti, older *te:k^T-énti, cf. 3sg
*té:k^T-ti in Ved. tá:s.t.i), therefore the -o- of *mon-éye-ti
should not have been there, or it should have taken the accent, and
I can see no reason for it to assume o-timbre.

The long-vowel counterpart is the type *swó:p-ie-ti 'makes sleep',
Lat. so:pi:re, ON sø:fa. Departing from root *swe:p- + suffix *-éye-
would lead to §*swépyeti (if pretonic long vowels are shortened,
remaining pretonic full vowels are then accented, and the de-
accentuated vowel finally lost, all in the normal course of events).
The actually established form *swó:pie- is an -o- longer than the
expected §*swép-ye-; and the same is seen with *mon-éye- in relation
to the expected stem §*mn-éye-. Both expectations are set on the
right track if an additional element -o- is simply inserted into the
form from the very beginning. That yields *m-o-n- directly, and *sw-
o-ep- will yield *swo:p- by contraction. Thus, the first result is
that we get things right if we simply conceive of the -o- as a
morphological element in its own right, i.e. something that has been
added to the root when the causative was derived. Since the -o- does
not influence the accent or the ablaut, it behaves just like a
sonant, i.e. an underlying consonant which only assumes syllabic
realization by virtue of its position in the surface form. That was
my first result: this -o- looked like a consonant, but I did not
believe *that*. But who was I to tell?

The second act was the "Saussure effect" as it is now called. Word
got around in the circle of Jochem Schindler's pupils that "o-grade"
had a deleting effect on laryngeals. This was in fact an old
observation hinted at in passing in a footnote by de Saussure in
1904 and elevated to certainty by Hirt who presented many examples
in his Indogermanische Grammatik, vol. II from 1921. Since Hirt did
not recognize the existence of laryngeals, it had sunk back into
oblivion until it was relaunched under Schindler's authority and
ascribed to him. The phenomenon was, however, given full attention
in Beekes' 1969 monograph on The Development of the PIE Laryngeals
in Greek. In his book Beekes arrived at the observation that
laryngeals do in fact vanish after o-vocalism, but he was reluctant
to ascribe any importance to it, since it did not make phonetic
sense to him. That is where I come in -

If we take a word like Greek pórne: 'prostitute', rather plainly
from an IE form *pórnaH2, derived from the root *perH2- 'trade' seen
in Gk. pérne:mi, aor. e-péra-ssa, OIr. renaid 'sells' (Celt. *pri-n-
a-), we miss a reflex of the root-final laryngal /H2/ in the
position before the suffixal /n/. The same goes for Doric
tólma: 'patience' (*telH2- 'support, tolerate'), tórmos 'hole'
(*terH1- 'bore'), Lith. spar~nas 'wing' (spìrti 'kick', *sperH-),
bal~sas 'voice' (bìlti, *bhelH-) and many others from all branches
that can show such a difference. Now, it would be very strange if a
laryngeal were to be deleted in a sequence like *porH2neH2,
*tolH2meH2 (no matter how accented), but not in the preform of Gk.
telamó:n, Skt. taliman- with e-vocalism. It would be even more
strange that the laryngeal is fully sounded after zero-grade, as in
IE *plH1-nó-s 'full'. Why would speakers who had no problems
pronouncing structures like *plHnó- and *pelHno- have to reduce
*polHno- to *polno-? That just does not make sense. HOWEVER - if
the -o- in question was an original consonant, it would have meant
additional complexity, and THEN it could be quite natural that
speakers accepting *plHno- and *pelHno- just did not accept *pXlHno-
, but simplified it to *pXlno, which then, LATER, developed into
*polno-. That was the second indication that the -o- here at work is
an original consonant.

Now things began to develop fast. The two showpieces pórne: and
tólma: share another peculiarity: They are both derived from shorter
stems with the suffix *-men-, which is reduced in both. The
reduction basically goes by a rule of dissimilation set forth by
Johs. Schmidt in 1895 (Kritik der Sonantentheorie). If the root
contains a labial the regular reduction product of *-mn- is *-n-,
otherwise it is *-m- (later analogy has strongly favoured *-m-, very
rarely *-n-, so that a phonetic rule is rather obvious). This
principled reduction is part of Sanskrit grammar in that stems in -
man- have reduced forms in the instrumental singular: mahi-n-á:,
vari-n.-á:, prathi-n-á:, pre-n.-á:, bhu:-n-á:, but dra:gh-m-á:, ras'-
m-á: . Now, why is this reduction restricted to the instr.? Because,
in a proterodynamic paradigm with basic alternation between the root
and the stem-suffix, the instrumental is unexpectedly accented on
the ending. This is seen in tu-stems which form such grammaticalized
structures as acc. kár-tu-m '(going somewhere) to do'; abl.
kártos 'after doing' (IE presumably *kWr-téw-s with zero-grade
root); dat. kar-táv-e 'to do' (likewise from *kWr-téw-ey); instr. kr-
tv-á: 'by doing' (from *kWr-tw-éH1). Now, in men-stems inflected
this way we have nom, -ma:, acc. -ma:n-am after original full-grade,
and weak cases in *-mén-s, *-mén-ey, *-mén-i with full-grade in the
suffix. In the instrumental, however, the ending *-eH1 would be
preceded by zero-grade of both root and suffix *-mn-, so here there
was once an initial sequence like *dlH1gh-mn- made up of consonants
only, and it stands to reason that -mn- was reduced here, but not in
*dléH1gh-mo:n or *dlH1gh-mén-s where the nasals are not contiguous,
nor in a neuter like bhú:ma, where the root was originally in the
full grade (*bhwáH-mn or *bhéw&-mn). That very strongly indicates
that the reduction of underlying /-men-/ via *-mn- to *-m- or *-n-
only occurs after VOWELLESS sequences. Ergo, the -o- of the mn-
reducing derivatives pórne:, tólma: could not be an old vowel. This
was the third piece of evidence demanding consonantal status of this
particular -o-.

But there is more. Up till now, I have not addressed the seemingly
embarrassing accent of pórne:, tólma: - why is it on an element that
was not even a vowel? A survey of the evidence that can be credited
with IE age yields the very clear picture that the accent is not
always on the root. It actually appears on the thematic suffix in
many forms - Gk. kormós, pho:né:, phlogmós, stolmós, but not in
others, like the mentioned *pórnaH2, *tólmaH2, *tórmos, *wólnos,
etc. A closer looks gives one a list of roots like *ker-, *bheH2-,
*bhleg-, *stel- for the first set, *perH2-, *telH2-, *terH1-,
*H2welH1- for the second. In other words: light roots form these
derivatives with accent on the suffix, heavy roots with accent on
the "root". The same distribtion can be shown for Balto-Slavic,
Germanic and Vedic and, especially, between the branches. Now, how
is this distribution to be explained? Well, if the -o- was
originally a consonant, it must have been syllabified at one time. I
can now simply assume that the pre-PIE accent shift to the first
full vowel of every word worked at a time when the infix consonant
had been syllabified after heavy roots, but not yet after light
roots. Then forms like the later *pornéH2, *tolméH2 would have two
syllables, while *kXrmós, *bhXH2néH2 had only one, and the accent
shift would lead to *pórneH2, *tólmeH2, *kXrmós, *bhXH2néH2, and
finally syllabification of the remaining X's would yield *pórneH2,
*tólmeH2, *kormós, *bhoH2néH2, whence Greek pórne:, tólma:, kormós,
pho:né:. I venture the statement that this could not have been
caused by an originally vocalic element. So that was the fourth
indication of the earlier consonantal nature of the -o-.

The fifth major pillar of the theory is the double-headed appearance
of the o-forms. At least the two types *mon-éye-ti and *tog-á-H2
have allomorphs without the -o-, appearing in derivatives from
special root structures. If causatives have causative function, they
always have the -o- in the root, but if they are used without the
causative meaning, then being called "iteratives", they may also
have plain zero-grade. Thus Vedic caus. roc-áya-ti 'makes shine' vs.
ruc-áya-te 'shines'. I guess the difference in diathesis is original
here, and that the iterative is simply the middle voice of the old
form of the causative: "makes oneself shine" or "is caused to shine"
has gone full circle and simply means pragmatically "shines". That
is the easy explanation why the middle voice is not obligatory
anymore. Now, beside *mon- we have *luk- in *mon-éye-ti, *luk-éye-
tor, and beside Old English mano 'mane' we have lufu 'love' from
*mon-á-H2, *lubh-á-H2. Thus, the two roots *men- go like each other,
and *lewk- goes like *lewbh-. We also have Latin toga beside fuga,
Gk. phugé: - and Gk. dokei^, Slavic dositi 'accept' beside Vedic
s'ubháyate 'adorns herself', meaning that *teg- goes like *dek^-,
and *bhewg- goes like *k^ewbh-. A full survey appears to show that a
given root structure has consistently -o- or consistently zero in
both of these formations. I have restricted my search to these
formations because they can be diagnosed with some confidence. The
conclusion is that the -o-, which is ABSENT in some root structures
(generally, but not entirely, the more complex ones), must have been
once a consonant that was either deleted because it caused too much
complexity or (in some special types) was assimilated to the
environment to such a degree that it escapes detection. That brings
the main basis of the consonantal o-infix theory up to five
independent observations.

There actually is more, for it can be shown that roots in initial /r-
/ do not infix, but prefix, the element. And so apparently do roots
of the structure *HeT-.

All of this demands that the /o/ concerned was once a consonant.
Now, that cannot be assumed for all IE o-vowels, especially not for
those that plainly are just accent variants of the e-vowels (as
*p&2té:r : *swéso:r). Now in fact, contrary to general doctrine, the
laryngeal-deleting effect is not to be ascribed to *every* IE /o/,
but *only* to the originally consonantal infix. This is seen from
well-established reconstructions with retained laryngeals after -o-
from other sources. Thus, the thematic vowel did not delete the -H1-
of *bhér-o-m&1no-s 'being carried'. The dissimilatory -o- of the
reduplicated intensive did not delete the laryngeal in e.g. *m(e)l-
mólH-ti 'grinds', cf. the acute of Lith. málti. And the acrostatic -
o- caused no laryngeal deletion in e.g. *twór&-s > pre-BSL *tvár&-s
> *tvá:r-s > *tva:~r- => Lith. tvorà, tvo~raN 'fence'.

The phonetic rationale of this is relatively clear: If the
consonantal predecessor of the later /o/ was a sound like a uvular
[R], it would be quite understandable that it did not survive before
most consonants, for r-sounds have a distinct predilection for being
last in initial clusters (English has tr-, str-, but not rt-, rst-,
srt-). Therefore, if fusion of two morphological elements had led to
the juxtaposition of something like [R] with a following word, the
natural consequence would be metathesis to move the r-sound into the
final position of the cluster. That brought it to the position
immediately preceding the root-vowel (*mRen-, *tReg-); when the
ablaut later deleted the root-vowel the infix came to occupy the
position of the root-vowel (*mRn- > *mon-, *tRg- > *tog-). That is
why it has been misinterpreted as a form *of* the root-vowel, which
I say it is not. If the word itself began with r- metathesis was of
no avail, so Rr- just remained and later gave *or-. For *o-HT-
instead of §*HoT- I can only guess: perhaps RH- was generally just
as good as HR- and so remained unchanged, or perhaps *HRT- did come
about, but was metathesised back again. The vocalisation product of
the consonant /R/ to an o-like vowel that eventually coalesced
completely with /o/ from other sources would have many parallels.

There remain two quite embarrassing loose ends: (1) If /R/ was once
just a phoneme of the language, we would expect it to appear in
other material than just this one morpheme, but where are the
examples? - (2) The functional service rendered by the infixal -o-
is not altogether clear. It creates adjectives, perhaps of
belonging, but exactly what is it doing in the stem of the
causative? My quick answer to (1) is that it may have coalesced with
some other phoneme in other positions, or have been lost, so that we
do not classify it as a separate element. Concerning (2) I would
point out that the problem remains if we call it "o-grade" or
even "a principled alternation of o-grade and zero-grade" instead.
Why does *that* turn up both in the causative-iterative and in the
toga-fuga formations? That strongly supports the analysis of these
o's as belonging together as manifestations of the same original
entity.

Jens