A Grammar of Tamashek (Tuareg of Mali)

70 3 Phonology

b. a-diiban

a-faffaru

a-jlwad

a-düban

'a-faffaru

'marriage'

'scrubbing'

'a-jiwad 'fleeing'

When the consonant(s) immediately following the vocalic prefix are
neutral (i.e., not BLC's, §3.1.2.2), the output of Prefix Reduction can be
diagnostic for the lexical representation of the first stem-internal V. This is
useful when this stem-internal V is phonetic [ε] or [o] before a BLC, since the
BLC merges preceding lexical {i e} as [ε] and preceding lexical {u o} as [o].
For example, the noun 'sunrise' is heard as [a'djmod], compatible with either
ά-jmud or d-jmod, since {u o} merge as ο before the BLC d. In most dialects
(including T-ka), but not A-grm, Prefix Reduction converts a- to a- before this
stem, which implies that native speakers take the [o] to be a high V (namely,
u). This test does not work in dialects like A-grm, which reduce a- and e- to as-
regardless of harmonic environment. Even in T-ka, the test fails to work when
both the prefixal V and the first stem-internal V are followed by BLC's, as in
[a'draz] 'sauce'. The reduced form is heard as [a'draz], a pronunciation that
could be phonemicized either as na-draz or "ae-draez. Based on the large
number of d-CCaC nouns whose vocalism is indisputable, I take 'sauce' to be
d-draz, but there is no obvious way to test this hypothesis.
For T-ka, Short-V Harmony also applies as a kind of filter to the first two
syllables of nouns and verbs. In other words, in T-ka, a but generally not ae is
possible in these contexts when the following syllable contains high {i a u}.
For verbs, this results in a large number of dialectal variants in perfective
stems such as T-ka -ammu-t 'die (PerfP)' versus -aemmu-t in most other
dialects (§7.3.2.1). In the case of verbs, A-grm sides with T-ka, though since
Short-V Harmony is not productive in A-grm it is best to account for the
schwa in this dialect as a function of vocalic ablaut melodies rather than as due
to a harmonic process as such. Since 3MaSg subject prefix i- combines with
schwa to form i but with ae to form ae, the dialectal differences are very
conspicuous in the 3MaSg subject form: T-ka (and A-grm) 1-mmu-t 'he died'
versus other dialects' 0-aemmu-t. A great many verbs are subject to this
dialectal variation in perfective verb stems; for example, the verbs of type
PerfP -akuf- 'swell, be inflated' (§7.3.1.9) and PerfP -affud- 'be thirsty'
(§7.3.1.10) have variants outside of T-ka and A-grm of the type -aekuf- and
-aeffud-. Clearly, Short-V Harmony is not particularly productive outside of

In the first syllable of a multisyllabic noun stem, T-ka usually enforces

harmony. This is seen in stems that appear elsewhere as e.g. -CaCi...,

disregarding any vocalic prefix. For example, 'fine sand' appears in most

dialects as e-daehi, but T-ka has e-dahi. However, I did record a few

exceptions in T-ka, such as t-e-säeni-t-t 'white spot above nose (e.g. of horse)'.

The lSg subject suffix -aer on verbs sometimes shifts to -ar- in the one

case where it is followed by a high V within the word (including clitics),

namely before directional clitics including Centripetal -\add (dialectally -\idd).

T-ka.