I have been asked to clarify my theory about tone assignment
on long vowels and diphthongs in PIE and Proto-Balto-Slavic,
so I will have another go at it. I hope all will be clearer
now.
The question I wanted to answer was: why do we have the
Greek acc.sg. a:-stem ending -é:n, with an acute, and the
gen.pl. ending -ô:n, with a circumflex, while in Lithuanian
we have a:-stem acc.sg. -aN and gen.pl. -uN, both with
circumflex intonation?
We can distinguish the following kinds of segments:
C - (consonants): no effect on the tone
V - (vowels): can have rising or falling intonation
R - (postvocalic or syllablic resonants: m, n, l, r, i, u):
always falling intonation
H - (postvocalic or syllabic laryngeals): always rising
intonation.
Long vowels can be either lengthened grade (V:) with rising
intonation (as if [aá]), or [PIE or PBS] contracted vowels
with falling intonation (as if [áa]). In Balto-Slavic, V:
merges with VH (after Hirt's law).
Simple diphthongs are always circumflex: V/R\
Long diphthongs can be:
V:R or VHR [circumflex]
An example would be the a:-stem acc.sg. *-ah2m. The
difference between Greek and Balto-Slavic is that in Greek,
m, n and r, l had passed into the C category, so that an
acute vowel á: followed by /n/ is not a falling diphthong
anymore, but an acute vowel followed by a consonant. In
Balto-Slavic, the cluster is still a diphthong, and the
diphthong as a whole has falling intonation (-a:/N\)
(despite the fact that the initial vocalic part is acute).
VVR (V::R) [circumflex]
An example would be the o- and a:-stem gen.pl. *-oom
(*-ô:m). In Greek, the vowel is circumflex, and the nasal
is irrelevant. In Balto-Slavic, the vowel is circumflex,
and so is the diphthong as a whole.
VRH [acute]
Short diphthongs followed by a laryngeal (further followed
by a consonant) acquire rising intonation in Balto-Slavic.
I see this as a matter of intonation only (V/R\ vs. V\R/):
the initial vocalic part is short.
A special case are the nominative singulars of r- and
n-stems, where we have PIE *-o:n (*-e:n) and *-e:r, i.e. a
lengthened grade (rising) vowel followed by a falling
resonant (and thus acute in Greek, circumflex in PIE). The
resonant could be dropped, but the falling intonation
associated with it remained, which results in the attested
endings -ô: (Lith. -uo~, Slavic -y) and -ê: (Lith. -e.~,
Slavic -i).
Exceptions to the general rule are:
1) In certain cases, -j behaves as a C, not an R. This is
the case in the o-stem nom.pl. -o-y (also in Greek, where
-oi always scans as a short syllable!), and in the
imperative *-o-yh1-, as opposed to the o-stem loc.sg. -o-i,
and the NAV n. dual -o-ih1, where j behaves normally as a
resonant.
2) Before word-final -s, all R's behave as C. This is the
case in the acc.pl. *-Vns (*-Vms) and in the o-stem ins.pl.
-o:ys.
This means that the /j/ or /N/ does not cause falling
intonation in these cases (although /j/ can do so
secondarily in Lithuanian).
Note that Proto-Slavic had only circumflex long vowels â:,
ê: and ô:. Circumflex î: and û: did not exist (they came
into existence after the monophthongization of *ei, and by
Meillet's law). This explains why secondarily
lengthened/duplicated *i and *u (e.g. sy"pjoN) are a.p. a,
while secondarily lengthened/duplicated *a and *e (e.g.
skac^jóN) are a.p. b (Dybo, SA, pp. 209-210).
=======================
Miguel Carrasquer Vidal
mcv@...