38 3 Phonologyfollowing V-initial subject suffix to form 3, in effect transferring its [+high]
feature to the surviving short V. The hi surfaces as a before C-initial subject
suffixes like 3FeSg -naet.
A good case can be made for a parallel underspecified and deletable low V
that I write IAI. This occurs only in long imperfectives of a subset (namely,
bisyllablic -vCCu-) of the same V-final verbs that have hi in the short
imperfective (preceding paragraph). The LoImpfP in question appears as
-PdQQ word-finally (i.e., with no subject suffix), as in -bass 'vomit'.
However, when a C-initial subject suffix is added, we get forms like 3FePl
bassas-nast with a stem-final low V, and this is the main evidence for a
representation of the type /-bdssA-/. When a subject suffix beginning in as is
added, we get forms like 3MaPl bdssae-n, which could be generated with or
without a stem-final /A/, but which I take (by parallelism to the short
imperfectives) as due to contraction, i.e. /bassA-asn/ with the IAI transferring
its [-high] feature to the surviving contracted V. In sum, the evidence for IAI is
weaker than that for hi, and IAI occurs in the paradigms of a subset of the
stems that have hi.
For more on these unaugmented V-final verbs see §7.3.1.3.
Both final hi and IAI in these verbs correspond to "u" in the corresponding
forms of Niger Tamajak (LTF2 424, class I.A.7-ll,e.g. Imprt "zeknu" and
LoImpfP "ikannu" for 'do well'). However, Foucauld's data do not have a
final V in such forms as "iläss" for 'he dresses' (DTF 4.2020) and "ed iken"
for 'he will do well' (DTF 2.821).
Synchronically in Tamashek, one might equate stem-final hi with a, and
IAI with ae, noting that {a ae} do not otherwise occur as stem-final segments.
However, these equations are not transparent, and the fact that the verbs in
question have stem-final full V's in the perfective and long imperfective stems
makes it difficult to justify short-vowel representations for hi and IAI. No
other stems or words in the language end with short V's.
If we were to identify hi and IAJ with full V's (the only V's that otherwise
occur in stem-final position), hi should probably be identified with i rather
than with u. This is because there are some verb classes with a non-deleting
final u, either in the imperfective forms only or in both perfective and
imperfective stems. For example, the verb 'cough' appears as -asu- in both
PerfP and Shlmpf stems (§7.3.1.4). This u does not delete word-finally, and
contracts with /ae/ at the beginning of a subject suffix to form u rather than a,
as in asu-n 'they-Ma coughed' from /asu-aen/. This true stem-final u therefore
has no phonological similarity to III. However, there is only one verb with true
i in the Shlmprt stem, namely -Iwi- 'be born'. Since this verb has a /-iCi-/
shape not found with any of the verbs that I analyse as having final hi in the
Shlmpf (and IAI in the LoImpfP), one could imagine identifying III with i and
trying to connect the phonological differences in some way with stem shape.
However, again this is a somewhat convoluted argument, and I prefer to think
of hi and IAJ as abstract segments with vowel-like properties.
The deletion rule may be formulated as (29).