A Grammar of Tamashek (Tuareg of Mali)

194 4 Nominal and pronominal morphology

b. agentives and related nominals

e-m-aert l-m-aert-an 'tree'

e-m-aeks l-m-aeks-αη 'eater'

e-m-aels i-m-aels-αη 'garment' (T-ka, A-grm)

e-m-aeqY "l-m-asijv-an 'killer'

e-m-aerz 1-m-asrz-an 'breaker' (T-ka, A-grm)

e-m-aesaeww l-m-assw-αη (T-ka) 'water source'

[PI also i-m-aesaeww-αη (A-grm)]

c. like (b) but nominals retain stem-final V in Sg

e-m-sesli l-m-aesl-αη 'voice'

e-m-aesli l-m-aesl-αη (A-grm) 'game, activity'

What is notable about the cases in (171.a-c) is that the PI has default
accent while the Sg has a grammatically marked accent on the final (which
corresponds to the penult of the suffixal PI). In effect, the Sg accent is erased
in the PI, allowing Default Accentuation to apply to the latter. This is quite
unusual in Sg/Pl pairs, whether the PI is suffixal or ablauted.
The verb stems associated with the nominalizations in (171.a-c) are of the
shape -aC(C)a- in the PerfP and /-aeC(C)i-/ in the Shlmpf. The underspecified
high V segment III is deleted word-finally by Stem-Final i/A-Deletion (29)
(§3.1.2.4), and is deleted by VV-Contraction (37.c) when followed by a
V-initial suffix or clitic (37, 40). In relative clause contexts, in those positions
where Rightward Accent Shift (132) (§3.5.3.2) would be expected to shift
accent onto the /i/, but where the III is (then) deleted, χ-Erasure (136) applies
(§3.5.3.3) and we get a default accent (if necessary, on the preceding word in
the accentual phrase).
Something related seems to be happening in (171). The verbal nouns in
(171.a) belong to types showing stem melody and with marked
penultimate accent, so we can take the Sg forms to have representations like
/a-tw-ajji/ 'being done', ending in underspecified high vowel hi. In the PI
/i-tw-ajji-aen/, the hi is lost by VV-Contraction (39.b). Even though the III is
not itself accented, its loss appears to trigger χ-Erasure. This produces
unaccented /i-tw-ajj-an/, which surfaces with default accent as l-tw-sjj-an.
The remaining cases in (171.a) have similar derivations. However, a-s-sksnn
'showing' shows some additional morphophonology in the Sg
(resyllabification by Final-CC Schwa-Epenthesis, accompanied by Stem-Final
Gemination) that does not apply to the PPi-s-alcn-an.
The agentives and related nominals in (171.b) can likewise be assigned
representations of the type /e-m-aerti/ 'tree' ending in III. The plurals are
derived in the same way as those of the verbal nouns, as shown in (172).