82 3 PhonologyI use ν (double grave accent) in two situations. First, certain suffixes and
clitics require that the default accent be on the (surface) penult rather than
antepenult of the word, effectively shifting the default location one syllable to
the right. Some of these suffixes and clitics may once have had an additional
vowel that has been lost. In this event, modern penultimate accent would have
once been antepenultimate and therefore compatible with Default
Accentuation. The best example of this is 3MaSgO clitic allomorph -\tt in
dialects like T-ka, since a syllabic variant -\tti is attested dialectally within
Tuareg. Another example, more difficult historically, is FeSg suffix -t on
nouns. Compare the accent of masculine a-bsembaera 'Bambara man' with that
of its feminine counterpart t-a-baembaera-t-t 'Bambara woman'. If FeSg -t
were a reflex of -tV, the original vowel would account (historically) for the
synchronic accentuation. However, I hasten to add that I know of no direct
evidence from other Tuareg varieties or other Berber languages that FeSg -t
was ever -tV. For a list of the relevant suffixes and clitics see §3.3.1.1, below.
Secondly, I use ν when a verb-stem-final V has been lost, and where an
underlying antepenult with default (not marked) accent ends up as an accented
penult. It should be noted that the interaction of VV-Contraction (37) with
Default Accentuation is complex (§3.3.2, below). For example, in the
PerfP/PerfN paradigm of -vnsu- 'excuse', the lSg subject form is anse-r,
while the 3MaPl subject form is ansas-n. The difference in accents is
observable when we prepose Neg waer, hence lSg war anse-γ Ί did not
excuse' with phrasal accent shifting to the preverb, but war anse-n 'they-Ma
did not excuse', where the accent remains fixed on the surface penult of the
verb form.
There is no phonetic difference among these accents. In phonetic
transcriptions I use [v^1 ] to indicate accent, regardless of its grammatical source.
In underlying transcriptions, only lexical and grammatical (=ablaut) accents
are indicated: /v/.
An accentual minimal pair is ae-jasu 'walking past (VblN)' and ae-jasu
'calabash'. There are many verbs whose PerfP (Perfective Positive) and Resit
(Resultative) stems differ only in accent, e.g. PerfP 0-üjaj 'he went far away'
and Resit 0-ujaj 'he has gone (=he is) far away'. This is because the Resit but
not the PerfP has an accent formative as part of its ablaut package. A more
exotic minimal pair involving verb stems is Imprt z-azar 'exhaust!' versus
Imprt z-azar 'put in front!'; the former shows resyllabification from -z-vzru-
(cf. dialectal variants z-azru, z-aezru), while the latter has a basic shape
-z-vzvr-.
Any word-form (stem plus any affixes and clitics) that can be pronounced
in isolation has a single primary accent on one of its final three syllables. If
there are two or more pretonic syllables, secondary accents (indicated in my
normal transcription by a grave accent 0) may appear on alternating syllables
counting back from the primary accent. Here I discuss accents within word-
forms, before turning to phrasal accents (which may override word accents),
see §3.3.3, below.