3.2 Local assimilations and syllabification rules 53deleted before the full vowel e, as we would expect. The Hortative suffix
requires accent on the preceding syllable, so we cannot really tell whether
VV-Contraction here precedes or follows Default Accentuation. Examples of
(37.a) are n-as-et-\add 'let's come!' (§7.2.3.3) with Shlmpf /-asi-/ 'come', and
long hortative na-jall-et 'let's go (every day)' (§7.2.5.5) with LoImpfP
/-jdllA-/.
Now consider cases where V 2 is a short vowel (37.b-d). In (37.b), the
suffixal /as/ is deleted after the full high V. We see this in asu-n 'they-Ma
coughed' and t-iwi-n 'they-Ma are born', both of which end in 3MaPl subject
suffix /-aen/. This treatment was observed in most Tamashek dialects, but I did
record as-aen 'they-Ma coughed' for A-grm, where the stem-final high V is
deleted before ae.
In (37.c), the stem ends in a deletable vowel, high Λ/ or low /A/. These
segments are deleted in word-final position, and show up as short a and as,
respectively, before a C-initial subject suffix. When they combine with suffix-
initial /ae/, the regular outputs are /i + ae/ —» 9, and /A + ae/ —» as. Since III is
arguably a special case of stem-final hi, and IAJ is arguably a special case of
stem-final /ae/, one could summarize these outputs as deletion of the suffixal
/ae/ following the stem-final short V. Examples, using 3MaPl -aen, are Future
ad £eks9-n 'they-Ma will eat' (/aeksi-asn/), and LoImpfP jallae-n 'they-Ma
(regularly) leave' (/jdllA-aen/). Because the quality of the stem-final V
determines the quality (ae or a) of the output, I put the hyphen after the output
V.
In the case of /i + as/, there are two suffixes that give the output ae instead
of the usual a. For these suffixes, therefore, /i + ae/ and IA + ae/ merge as ae.
The first suffix is MaPl Imprt O-aet. The effect is that C)-aet is just added to
the unsuffixed Sg imperative (prior to ^syllabification if applicable). Thus
Imprt aeks 'eat!', MaPl Imprt asks-aet 'eat!', with /aeksi/. It is questionable
whether seks-aet is really produced by VV-Contraction applying to /aeksi-aet/; it
may be more realistic to first derive Sg Imprt aeks and then add the MaPl Imprt
suffix as an outer morphological layer.
The other problematic suffix is lSg -aer. Because r is a BLC, we could
get phonetic -aer from either /asr/ or /ar/, so we might allow VV-Contraction
to apply in the same way as for e.g. 3MaPl -aen and then have a late rule merge
ae and a into ae before y. In this way, VV-Contraction could apply in the same
way to lSg -aer as it does to the other (non-imperative) subject suffixes
beginning in ae, though the distinction between ae and a outputs would later be
neutralized. However, this analysis will not work.
Consider the inflectional paradigm of stems like Shlmpf /asksi/ 'eat': lSg
asks-asr, 2Sg t-aksa-d, 2MaPl t-aksa-m, 3MaPl aksa-n. We may throw in
2FePl Imprt aksa-maet for good measure. Note that all the combinations except
lSg aeks-aer show a surface a in the first (as well as second) syllable. This is
due to Short-V Harmony (§3.2.6), where the initial syllable shifts /ae/ to a
under the influence of a in the following syllable. The fact that this does not
happen in lSg äsks-asr shows that the second syllable has /ae/ rather than /a/ at