54 3 Phonology
the point when Short-V Harmony applies. Therefore, VV-Contraction must
exceptionally convert /i + as/ to as instead of a in connection with the ISg
suffix. For this reason, I transcribe aeks-asr with the hyphen before the suffixal
as.
Since, before a C, stem-final III is realized as a, while I Al is realized as as,
one could simply equate hi with a and I Al with as. Or we could just say that
these abstract segments "become" a and ae, respectively, before a suffix,
including a V-initial suffix. If so, the regular formulae h + as/ —> a and /A + ae/
—> ae in (37.c) reduce to /V, + as/ Vb exactly parallel to the cases with V, =
full high V in (37.b). Can we do the same when V, is /a/, followed by a short
suffixal V?
In (37.d), we see variable treatment of la + ae/. There are actually three
surface outputs of this combination, namely as, e, and a. However, the a output
and some instances of the e output are secondarily derived from /ae/, by belated
attachment of ablaut formatives.
Consider the following 3MaPl forms for 'eat': PerfP aksae-n, PerfN
akse-n, and Resit aksd-n. These forms reflect the ability of verb stems of the
(light) shape -v(C)Ci> to include the material up to and including the first C
of a subject suffix in the domain of ablaut. This rebracketing allows the
entire 3MaPl suffix -aen to be included in this domain; for 3FePl -naet the
domain stops at the n. When combined with -vksu- 'eat', the 3MaPl PerfP
/aksa-aen/ is realized as Sksae-n, showing an apparent contraction la + as/ to
short ae. The sequence sksae-n, when under negation, is subject to PerfN ablaut
(formative e-pclf, §7.2.2.3), which changes /ae/ in the relevant position (first
postconsonantal vowel, also final-syllable vowel) to e. Likewise, aksas-n can
take Resit ablaut (χ-pcl and χ-pcl, §7.2.2.2), which lengthens and accents the
first postconsonantal vowel, turning /ae / into α. In this analysis,
VV-Contraction itself is not responsible for the e or α outputs, just for the
initial as output. Behaving like 3MaPl -aen in these respects is 2MaPl -asm.
After heavy (though not light) V-final non-augment verbs, we can add ISg
-asir and 2Sg -aed (to make a clean sweep of V-initial subject suffixes). Thus
-rvftu- 'have a scare', 3MaPl PerfP arrsftas-n, ISg arraft-aer.
However, the short output ae seems rather odd phonologically for la + ae/,
i.e. for the combination of a full and a short V. I will now argue that the
regular phonological output for this sequence (in verbs at any rate) is not ae but
e.
In fact we get e rather than ae when any /aeC/ subject suffix, including ISg
-aev, is added to an augment verb. Verbs of the augment class end in a full V,
as seen most clearly in their VblN (which end in i or u). The combination of
inflected verb plus C-initial subject suffix, or zero suffix (in connection with a
subject prefix), requires Augment -t- immediately following the verb stem.
However, a V-initial subject suffix such as 3MaPl -aen or ISg -aer does not
allow the Augment, so VV-Contraction must occur. The result is e. Example
('be in large quantity'): VblN ά-baffu, 3MaSg PerfP I-bbuffe-t (T-ka,