The origin of the Slavic o-stem Npl. ending -i has been much
debated. The traditional account, as found for instance in
my old University textbook (Bräuer's "Slavische
Sprachwissenschaft"), and ultimately due to Fortunatov,
claims that PIE final *-oi and *-ai developed to -ê when
circumflex, but to -i when acute. This hardly seems adequate
to explain the evidence, which shows -i in cases where no
source for an acute is likely (the o-stem N.pl. -i < *-oi,
for instance), and -ê in cases where we would expect an
acute (e.g. the a:-stem NA du. -ê < *-oiH).
I have hinted at my solution to this problem here before,
but I think the issue deserves fuller treatment.
We first turn to Classical Greek accent laws, specifically
to the "law of limitation". As described by Philomen Probert
("A New Short Guide to the Accentuation of Ancient Greek"):
> (a) An acute accent may not fall further from the end of the
> word than the antepenultimate syllable (legómenos is possible
> but *légomenos is not).
> (b) A circumflex may not fall further from the end of the word
> then the penultimate syllable (so:tê:ra is possible but *sô:te:ra
> is not).
> (c) If the final syllable contains a long vowel or ends with a
> consonant cluster [], no accent may fall further from the end of
> the word than teh penultimate syllable (legoménou, pomphólux,
> and polupîdax are possible, but *legómenou, *pómpholux,
> and *polúpi:dax are not).
> (d) If the final syllable contains a long vowel, a circumflex may
> fall only on the final syllable (liguroû and anthró:pou are
> possible but *anthrô:pou is not).
> ["the so:tê:ra rule"] If the final syllable contains a short vowel
> and the penultimate syllable contains a long accented vowel, the
> accent on that vowel _must_ be a circumflex (so:tê:ra is possible
> but so:té:ra is not).
Now rule (c) mentions long/heavy final syllables, and there
is one further consideration here. I quote:
> In general, diphthongs count as long vowels for the purposes of
> the accent. The diphthongs -ai and -oi, however, count as short
> for accentuation when they occur at the absolute end of a word in
> indicatives, subjunctives, imperatives, infinitives, or nominatives
> plural: boúlomai, boúlontai, boúlo:mai, boúlo:ntai, paídeusai
> (imperat.), paideu~sai (aor. inf.), ánthro:poi, trápezai, oi~koi
> (nom.pl.), boulómenoi, boulómenai.
> But in the optative final -ai and -oi count as long: paideúsai (opt.),
> paideúoi. Final -ai and -oi likewise count as long in adverbs with
> the meaning 'at (a place)', e.g. oíkoi 'at home', Megaroi~ 'at Megara',
> and in certain interjections, e.g. aiai~ 'alas'.
> Notice that although some final diphthongs count as 'short' for the
> purposes of the accent rules, they make a syllable 'scan long' in
> poetry just like other diphthongs.
On the basis of the Greek distribution ("short" N.pl. -oi
vs. "long" L.sg. -oi), we may put forward the hypothesis
that the distribution of final -i and -ê in Slavic goes back
to a PIE distinction between *-oj with consonantal /j/ [-VC]
vs. *-oi with vocalic/glide /i/ [-VR].
Apart from the o-stem L.sg. (*-oi) and N.pl. (*-oj), there
are some other forms which need to be considered:
- the optative *-oih1- (Slavic -i => *-ojh1)
- the NA n./f. dual *-oih1 (Slavic -ê => *-oih1)
- the personal pronoun datives mi, ti, si (=> *moj, *toj,
*soj)
- the 1sg. present (< perfect) vêdê (=> *-h2ai)
Not usually considered in this context, but worth taking a
look at:
- the 2sg. athematic ending -si (if from *-saj)
- the infinitive ending -ti (if from *-taj)
At first sight, the optative seems to speak against the
hypothesis: we have Greek -oi which is long for the purposes
of the limitation rule vs. Slavic 2/3sg. optative -i, as if
from *-ojh1s/*-ojh1t. But the Greek limitation rule
considers long final _syllables_, and an ending *-ojh1t
(-VCCC) must surely count as long (for that matter, so would
*-oih1t = -VRCC if laryngeal loss was late in Greek).
The dual ending *-oih1 has, to my knowledge, not survived in
Greek.
The Greek forms emoí and soí have an acute, but the
reflexive is hoî. I don't know what to make of that.
Probably nothing.
The Greek middle endings -mai, -sai etc. count as short,
which would seem to be in disagreement with Slavic vêdê, but
vêdê is _not_ a middle form: it is a perfect (Grk. oîda)
with -i added (as in Latin pf. 1sg. -i: < *-h2a-i), and that
-i may or may not be identical with middle -i. But the 2sg.
middle ending *-sai (analogical replacement of older
*-th2ai) does seem to be a good candidate for the mysterious
Slavic 2sg. athematic ending -si (and the Lithuanian 2sg.
-ì). Lithuanian -ì can come from *-i, *-iH or *-eiH, Slavic
-i can come from *-iH, *-eiH or *-ei, but there simply
aren't any suitable endings *-iH or *-eiH that could have
supplied a 2sg. present ending. Under the theory I'm
putting forward here, Slavic -i and Lithuanian -ì may also
reflect PIE *-aj and *-ajH. If Lithuanian -ì results from
reanalysis of *es-saj as *es-aj, then the Balto-Slavic forms
may reflect the 2sg. middle ending *-saj (Greek "short"
-[s]ai).
Another possibility would be the 2sg. optative *-ojh1s, with
reanalysis this time in Slavic (*es-ojh1s => *es-sojh1s),
which would make the Slavic 2sg. athematic cognate with the
2sg. imperative. However, it does seem strange that a
thematic optative ending would become the marker of the 2sg.
athematic. And as far as I can see, Lithuanian would not
have dropped final -s in 2sg. *-ojh1s.
Which brings me to the Lithuanian optative.
At the "A. Richard Diebold Center for Indo-European Language
and Culture", Baltic, Old Lithuanian, Verbal system
(
http://www.utexas.edu/cola/centers/lrc/eieol/litol-7-X.html#Lit07_GP34_01),
I find the following information:
>..Lithuanian has forms deriving from the Indo-European optative. The East Lithuanian Vilnius
>dialect has retained forms the basis of which is not the infinitive, but rather the present
>tense stem, e.g.,
>
>2nd sg nès^(i) 'carry' (3 pres. ne~s^a '(he) carries'), duõd(i) 'give' (3 pres. dúoda '(he) gives').
>2nd sg mýly 'love' (3 pres. mýli '(he) loves'), z^iu:ry~ 'look' (3 pres. z^iu:~ri '(he) looks').
>2nd sg dãrai 'make' (3 pres. dãro ' (he) makes'), gáudai 'catch' (3 pres. gáudo '(he) catches').
>
>This type of the imperative mood is related to the desiderative mood (permisive). It has only the
>third person forms derived from the Indo-European optative. They are reinforced with the preposed
>particle te-, e.g.:
>
>tenes^ie~ 'let (someone) carry';
>tez^iu:ry~ 'let (someone) look';
>tegáudai 'let (someone) catch'.
The Vilnius forms cited above are 2sg., but lack final -s.
They are not, however, identical to the 2sg. indicative
forms in their accentuation. What am I to make of these
forms?
The infinitive presents a clearer set of correspondences. We
have Slavic -ti, Lithuanian -ti and shortness in the Greek
infinitives (which, on the other hand, are not at first
sight etymologically related to the Balto-Slavic ending: the
relevant Greek endings are -nai, -sai, -sthai). We might
posit an ending *-taj, of unclear origin, instead of the
usual reconstruction, which derives the Balto-Slavic
infinitive from the L.sg. *-te:i or D.sg. *-tejei of an
i-stem verbal noun in *-tis. Slavic -ti can be derived from
such forms, but not, I think, Lithuanian -ti (one would
expect *-tie~).
If we compare Lithuanian with the Slavic (and Greek) forms,
we have:
- the Npl.
Lithuanian nominal -ai(~), adjectival -ì. The nominal ending
is perhaps from the stems in *-ijos, which have -iai even in
the adjectives (from unstressed contracted *-ijaj).
Adjectival -ì is as expected from PIE *-oj. Pronominal tie~
with metatony in a monosyllable?
- the Lsg.
Lithuanian (as Greek) has preserved this ending only in
adverbial fossils such as namie~ 'at home' (why is that
end-stressed?). Otherwise as expected from *-oi.
- the optative
Was discussed above. We have -i, -ie~ and -ai. Needs
clarification.
- the NA n./f. dual
We have a:-stem -ì, which if from *-oih1 (c.q. *-oih2),
would indicate that, unlike in Slavic, the acute/circumflex
distinction _does_ play a role in the Lithuanian reflexes of
*-oj(H) vs. *-oi(H).
- the personal pronoun datives *moj, *toj, *soj
Not attested in Lithuanian?
- the 1sg. present (< perfect) vêdê (=> *-h2ai)
Not attested in Lithuanian.
- the 2sg. athematic ending -si (if from *-saj)
Lithuanian -ì, possible reanalysis discussed above. Extended
to all 2sg. endings. Does the 2sg. trigger Saussure's law,
or does the accentuation simply reflect that of PIE *-sái?
- the infinitive ending -ti (if from *-taj)
Lithuanian -i, discussed above.
=======================
Miguel Carrasquer Vidal
miguelc@...