As noticed by Dybo, the distribution of je-verbs with
lengthened root vowel is the following:
- i > i:, u > u: are in a.p. a (sy"pati, my"kati, smy"kati,
ty"kati, sy"sati, pry"skati, bry"zgati, sti"gati)
- e > e:, a > a: are in a.p. b (skaka"ti, xapa"ti, xrama"ti,
maka"ti, kaza"ti, drêma"ti)
The lengthening of the root vowel in these verbs must
therefore be relatively ancient, as it follows the PIE
distribution where the only long /i:/ and /u:/ were acute
(from /iH/ and /uH/), while a:, e: and o: could be either
acute or circumflex. In particular, the formation of these
verbs must predate the development *ei~ > i~ and Meillet's
law (which created new i~'s and u~'s (y~'s)).
The lengthenings seen in bIra"ti > bira"ti or zUva"ti >
zyva"ti are more recent: I and U lengthen to circumflex (or
at least non-stress attracting) i and y, and the relative
youth of these formations is confirmed by the fact that
these are aje-verbs, not je-verbs.
There is a category of verbs with lengthened root vowel
which could be even older than the group of
sy"pati/sti"gati/skaka"ti/drêma"ti, at least it is claimed
to be of already PIE origin by the makers of LIV. These are
cauative/iteratives (Slavic i/i-verbs) with a lengthened
root vowel, LIV category 4b (R(ó:)-je-).
If the claim were false, and these formations were of early
(Balto-)Slavic age, one would expect these verbs (all with
root vowel /a/) to fall into a.p. b, like the
skaka"ti-verbs. This is not the case. Instead, we find the
verbs in question scattered over all three accent paradigms:
ga"ziti, pa"riti and va"diti are a.p. a
davi"ti, pali"ti and travi"ti are a.p. b
sadi"ti and gasi"ti are a.p. c.
I'm not sure about the a.p. of kaniti, rac^iti and mariti,
but they appear to be a.p. b or c (SCr. kániti kâni:m,
máriti, mâri"m, Russ. rac^ítel'nyj). I'm ignoring skaka"ti,
kaza"ti and ka"jati, which LIV puts into this category,
because they are not i-verbs in Slavic (even though ka"jati
may well be an original R(ó:)-je-verb).
What explains this distribution? The verbs are not
uniformly a.p. b, so they cannot be lengthened /a/'s of the
skakati-type (/aa/ > /a~/). But a lengthened vowel of PIE
origin (Dehnstufe) should be reflected in Balto-Slavic as an
acute long vowel (/a:/ > /a"/), and the verbs are not
uniformly a.p. a either. Could the length be (Balto-)Slavic
(and therefore circumflex) and the exceptions with an acute
be due to laryngeals in the root? Most definitely not. In
fact, quite the opposite is true: all the verbs with a
laryngeal in the root are a.p. b or c! So perhaps that is
the pattern: roots of the structure /o:C/, /o:R/ are a.p. a,
while roots of the structure /o:RH/ are a.p. b. We have:
*g(W)o:g^h-éje- => gó:z-i:-
*(s)po:r-éje- => pó:r-i:-
*wo:dhh1-éje- => wó:d-i:-
as opposed to:
*dho:uH-éje- => do:w-í:-
*tro:uH-éje- => tro:w-í:-
*ko:nh1-éje- => ko:n-í:-
This now makes perfect sense. In the first set, the
syllabification is *g(W)o:-g^hé-je-, *(s)po:-ré-je-,
*wo:-dh(h1)é-je- with a long rising vowel in the first
syllable. In the second set, it is *dho:u-Hé-je-,
*tro:u-Hé-je-, *ko:n-h1é-je-, with a long falling diphthong
in the first syllable, and therefore a circumflex (cf. the
circumflex in *o:u-yóm > o~jé > vâje/jâje). When the
laryngeal fell away, the semivowel/resonant was pulled to
the next syllable, but the circumflex accentuation stayed.
The rule apparently does not apply to the causatives saditi
and gasiti (a.p. c), but it may apply to the causative
paliti (a.p. b). The root is reconstructed as *pel- in LIV
("Nur slav., Anit.-Wurzel, vgl. russ. pólomja "Flamme" (skr.
pla``me:n zeigt sekundäre Kürzung"), but if <polmja>
(*polHmn?) was mobile, Meillet's law would have taken care
of the acute. The other two exceptions are doubtful
(rac^iti (*rek-) may be a Germanic loanword, and of mariti
(*(s)mer-) the footnote in LIV says that its "Zugehörigkeit
[ist] zweifelhaft").
In short, the accentological evidence from Slavic seems to
suggest that iteratives with a long root vowel are indeed a
category inherited from PIE. Furthermore, it confirms that
PIE long non-circumflex vowels give a Balto-Slavic acute
(unless followed by a semivowel/resonant, but we knew that
already: Lith. s^uo~ < *k^wo:n, Lith. ah2-stem acc.sg. -aN~
< *-a:m < *-ah2m).
=======================
Miguel Carrasquer Vidal
mcv@...