A Grammar of Tamashek (Tuareg of Mali)

344 7 Verbal morphology

Prohib), as is true of long imperfectives in all stem classes. If the optional -t- in

-(t-)aPPaC- is disregarded, we can save the generalization that the accent

targets the first postconsonantal V ("pel") in all three cases. We also observe

lengthening of the accented V in all three cases, from /ae/ to ά. It appears,

however, that -PdQQaeC- and -t-aPPaeC- show lengthening due to the χ-pcl

component, which targets the first postconsonantal V; as usual with this

component, it applies only to the LoImpfP (and is absent from the LoImpfN

and Prohib stems). On the other hand, the lengthening in LoImpfP -(t-)aPPaC-

for -vPvC- stems appears to be due to χ-f (lengthening of the V of the final

syllable), since (as with χ-f in trisyllabic or longer stems) it is observed in

LoImpfN -(t-)aPPiC- and Prohib -(t-)aPPaC- as well as in the LoImpfP.

Since the Prohib (like the LoImpfP) has a melody or in all

three types, the overlain stem-wide of the LoImpfN is distinctive.

The three verb types have sharply different productive VblN patterns:

a-PseQaC, uPsC, and e-PeC. The prefixes in a-PsQaC and e-PeC are subject

to Prefix Reduction in the dependent state (§3.5.1), and have MaPl suffixal

plurals with i-...-aen. The uPaC VblN type has a productive PI üPCaw-aen in

T-ka (also attested in R and K-d). A-grm usually has uPaC-asn or uPC-an.

This uPsC looks quite isolated in the context of (347-8), but the same -vPPvC-

verbs also show -uPvC- shapes in prefixal derivatives such as the causative

and agentive. For example, -vqqvl- 'go back' has a causative -s-urvl- 'give

back' with Imprt s-üral, while -vddvr- 'be alive' has an agentive a-m-udaer

(§8.8.1).

The alternation between -vPPvC- and derived -üPvC- gives credence to

Prasse's view (MGT 6.69-70) that the verbs in question reconstruct as

*-vwPvC-, the semivowel sometimes fusing with Ρ to form geminated PP, and

sometimes fusing with the first ν to form u. In further support of this is the fact

that no #-vwPvC- verbs with surface w occur.

Prasse (MGT 6.72-3) also argues, less convincingly, that the -vPvC- type

(exemplified above by -vwvt- 'hit') reconstructs as *-vPhvC-. While this

would result in a certain analytical unification, whereby all three types

reconstruct as *-vPQvC-, I see no evidence (direct or indirect) for the putative

*h. Synchronically this analysis is challenged by verbs like -vdhvl- 'help'.

The three verb types are quite common. A few further examples (cited in

the Imprt) follow. For the -vPQvC- type: abdad 'stand', ablaj 'cross', absay

[a'bsi] 'melt', adhal 'help', adlal 'dance', adlam [adlam] 'harm, ardar

[ae'rdasr] 'betray', ajrah [a'jraeh] 'understand', and ajraw [a'jru] 'get'. Note

that the second and third C's may be identical (this has no morphological

consequence).

For -vPPvC-: addah [a'd:aeh] 'pound (in mortar)', affar [a'fiasr] 'hide',

affay [a'f:i] 'pour', and ammar 'look for'. I know of no -vPPvC- verb with

geminate ww (original *ww would presumably be reflected as gg). Verbs with

perfective -sewwaeC- are variants of the -uCvC- stem class, see §7.3.1.6.