122 3 Phonology(95) Derivation of PerfP -aexbabas-t '(hole) gape' (T-ka dialect)
/-xvbubu-t/ underlying
/-axababas-t/ various early rules and <H L> melody
/-axbabae-t/ Stem-Initial Syncope
/-aexbabae-t/ Leftward L-Spreading
-sexbabae-t surface formBecause Melodic Association for perfective verbs produces sequences of
the maximal type «HHLLL», there are never more than two Η-vowel syllables
at the beginning of the stem. In all relevant cases, the perfective stem begins
with a nonlexical initial short V, the first schwa of /axababae-t/, which is
therefore followed by no more than one other schwa-syllable before the
melody switches from Η to L. Stem-Initial Syncope deletes this second schwa,
giving /axbabas-t/. It remains merely to change the initial /a/ to as. If we
allowed Short-V Harmony to work in a symmetrical fashion, we could
attribute the initial ae of -aexbabae-t to this harmony rule. However, sequences
like aCCa are stable in other contexts, including perfectives of -vCCu- verbs
like -aksa- 'eat', so I have elsewhere expressed doubts about a symmetrical
formulation of Short-V Harmony (§3.2.6). As a result, I recognize Leftward
L-Spreading as a morphologically sprecialized rule converting the initial /a/ to
ae in (95).
There are some alternative options, especially for dialects other than T-ka.
In T-ka, the maximal melody for perfectives is «HHLLL», and there is no
reason to consider an alternative «LHLLL» which would unproblematically
become «LLLL» after Stem-Initial Syncope (deleting the Η vowel). However,
in the dialects other than T-ka and A-grm, it is much more reasonable to allow
«LHLLL» as a maximal melody for perfectives. In particular, in these dialects
(K-d, R, T-md), we get surface «LHLLL» when the Η syllable is full, i.e. i or
u. This applies even to verbs like -xvbubu- (+ -t), namely in the Resit stem,
which appears as -axlbabas-t in T-Ka but as -aexlbabae-t in the other dialects
mentioned. In these non-T-ka dialects, it may be best to formulate Stem-Initial
V-Insertion (see below) so that the added stem-initial V is specified as as. This
would actually work for T-ka as well, since Short-V Harmony would convert
/ae/ to a under the influence of the Η vowel of the following syllable. However,
in T-ka there is no direct evidence for initial /as/ in these perfectives, and since
«HHLLL» is a reasonable mapping of
inclined to have Stem-Initial V-Insertion just insert an unspecified short V in
T-ka.
There is an intriguing similarity between Leftward L-Spreading in (95) on
the one hand, and a dialectal alternation of the type -aPaQ- versus -aePQ- in
the short imperfective (including Imprt). Examples: Future ad t-awat 'she will
hit' versus, depending on dialect, ad awat-aer (e.g. T-ka) or ad aewt-aer (e.g.
K-d) Ί will hit'. The variants of type -aewt- occur chiefly in the Kidal area
((49), §3.2.6.2). These variants show the same pattern suggested in the