A Grammar of Tamashek (Tuareg of Mali)

80 3 Phonology

The suggested derivation of the PerfP stems is (59). This assumes a prior

application of Melodic Association, whereby <H...L> produces a sequence

«HHLL L» when mapped onto a pentasyllable string (§3.4.5).

(59) Derivation of -aenhaeltaettaew-

/-snshaeltaettaew-/ after Melodic Association

/-anhaeltasttaew-/ Syncope

-aenhaeltaettasw- Leftward L-Spreading

The spreading rule can be formulated as (60).

(60) Leftward L-Spreading

when Syncope applies after the initial C of a (perfective) verb, the

L component of the melody spreads across the site of

Syncope to the preceding (i.e. stem-initial) syllable, effectively

erasing the Η component of the melody.

If Melodic Association is formulated in such a way that perfective stems

never have more than two initial Η syllables, since the syncopated V is always

the s of the second syllable, the spreading is always confined to the

immediately preceding syllable (the stem-initial syllable). The fact that the Η

component is entirely erased may therefore be an "accidental" effect of a rule

that directly targets a single syllable.

Because it is closely associated with Syncope, it is difficult to discern a

(local) phonological logic to Leftward L-Spreading. For example, the rule

applies to post-Syncope PerfP /-anhaeltaettasw-/ in (59), but does not apply to

(unsyncopated) PerfP -ahlaek- 'destroy', which also begins in -aCCaeC-. As a

result, I regard Leftward L-Spreading as a morphophonological rule.

It is possible that Leftward L-Spreading as formulated here, making

specific reference to Syncope, is historically incorrect. In the A-grm dialect,

the relevant perfective stems show the same stem-wide surface melody as

in T-ka and the other dialects, but in A-grm Syncope does not apply here. For

example, in A-grm we get PerfP -nashaeltaettasy- 'sway' and -xsebabas-t '(egg)

be hollowed', cf. T-ka -aenhseltaettaew- 'sway' and -aexbobae-t '(hole) gape',

cited above.

Whatever the historical origin may be, speakers of most Tamashek dialects

have little exposure to A-grm speech, and data from this peripheral dialect do

not suffice to shoot down (60) as a synchronic formulation for K-d, R, and

T-ka varieties.]