PIE Word Formation (4)

Long vowels in PIE. Static accentuation

Underlyingly short full vowels were lengthened in PIE in the final
syllable of consonantal stems in the nominative singular of animate
nouns (with the ending *-s) and the collective (with the ending *-h2).
The ending may be deleted after certain stem-final consonants (e.g. *-s
is lost after *r, *l, *n, *s and possibly *j and *m) or retained, but
the lengthened vowel is present in either case:

*ph2tér-[nom.sg.] --> *p&2té:r 'father'
*djéw-[nom.sg.] --> *djé:us 'sky, the Sky God'
*nép(o)t-[nom.sg.] --> *népo:ts 'grandson'
*wéd(o)r-[coll.] --> *wédo:r 'water(s)'

In the last two examples the *o reflects a post-tonically reduced vowel
which survives in the cited forms precisely because it was lengthened
before the operation of sound-changes deleting unaccented short vowels.
(Short *o was introduced analogically e.g. in the acc.sg. *népotm., so
that "strong" cases with *nepot- are distinguished from "weak" cases
with *nept-.) There is no lengthening if the vowel is directly followed
by the ending without an intervening consonant, as in *wl.'kW-o-s
'wolf', *mn.-tí-s 'thought, mind', *pért-u-s 'ford' or *bHor-á-h2 '[act
of] carrying'. Before a stem-final consonant cluster the vowel is short,
e.g. *h1s-ént-s 'being', presumably because of secondary re-shortening:
the participial suffix *-e/ont- has the old by-form *-e/on-, whose vowel
_is_ lengthened in the nominative singular.

These instances of lengthened vowels are only partly conditioned by the
phonetic context in which they occur; the morphological environment is
also relevant. It is interesting to note that while the *-s of the
nom.sg. has a lengthening effect, other homophonous endings, in
particular the *-s of the second person singular (in verbs) and the *-s
allomorph of the genitive singular, display no such influence. This may
indicate the nom.sg. *-s derives from a different pre-PIE phoneme than
other kinds of *-s endings.

Apart from this kind of derived length, PIE also seems to have had
_underlying_ (lexically specified) long vowels in some roots. These are
best visible in the verb system: the subtype of root verb known as
"Narten presents" (after Johanna Narten, who demonstrated the existence
of the pattern in question) has forms such as 3sg. sté:u-ti / 3pl.
stéw-n.ti 'praise' vs. the usual type represented by
*gWHén-ti/*gWHn-énti 'strike, kill'. The 3pl. ending *-enti, when added
to a consonantal stem, normally has a full vowel and attracts the
accent, causing phonetic reduction in the root. If *steu- is a reduced
(though still full-vowelled) variant of *ste:u-, the 3pl. ending must
have been accented at some point in the prehistory of PIE, but the
retained full vowel in the initial syllable drew the accent from the ending:

pre-PIE *ste:w-énti > *stew-énti > *stéw-enti > PIE *stéw-n.ti

A similar pattern occurs in the "sigmatic aorist": 3sg. *wé:g^H-s-t,
3pl. *wég^H-s-n.t . It is not quite clear whether the lengthening is
connected directly with the *-s- extension. My own view is that it
isn't, and that the *-s- (a morphological marker of punctuality) was
redundantly suffixed to root aorists already containing a lengthened
vowel (like Narten presents). All this, however, is still a moot issue.

The same root may have forms with and without Narten length. The
variation is not regulated by phonological factors and seems to be
purely morphological. It is thinkable that the lengthening was originaly
iconic and had a function similar to that of reduplication.

Traces of Narten ablaut possibly occur in PIE substantives and
adjectives as well, disguised as *o/*e variation in nouns like *pó:d-s,
acc. *pod-m., gen. ped-ós 'foot'. If accentual mobility in this type is
secondary (cf. lexically isolated archaisms such as *nókWt-s, gen.
nékWt-s), the o-grade forms can be seen as transformed reflexes of an
original lengthened grade. Jens Rasmussen argues that the o-timbre arose
in the nominative singular, where the underlying long vocalism of the
root plus the lengthening effect of *-s produced trimoraic overlength,
realised as a "drawled" vowel with two peaks of prominence, and the
second, unaccented peak developed an o-timbre:

*pé:d-[nom.sg.] --> *pé::d-s > *péo:d-s (or the like) > *pó:d-s

The other "strong" cases were analogically replaced with o-grade forms
(acc. *pód-m.). The "weak" cases kept their *e but not always their
accent, which in most cases was shifted to the ending on the analogy of
the "plain" athematic type (mobile *ped-ós for older static *péd-s in
the nominative singular). As elsewhere, a long vowel gets shortened if
followed by a stem-final consonant cluster: *nókWt-s rather than *no:kWt-s .

The alternation of *e:/*e can be discerned in *jé:kW-r. 'liver', gen.
*jekW-n-ós (for static *jékW-n-s). The type of ablaut represented by
nom./acc. *dór-u, gen. *dér-u-s (restructured as *dr-óu-s, *dor-w-ós,
etc.) is more mysterious, but must be somehow derivable from the Narten
pattern, whatever the source of the extra length that underlies the *ó
of the nominative/accusative (shortened before stem-final */rw/).

An original long vowel which never surfaces as such can be reconstructed
in static nouns of the type *stéu-to:r 'one who praises', gen.
*stéu-tr.-s (with later restructurings). Here the accent was once
shifted to the suffix or, in the weak cases, to the declensional ending,
and then drawn back to the root syllable, whose vowel had meawhile been
shortened but not reduced to zero.

Piotr