A Grammar of Tamashek (Tuareg of Mali)

58 3 Phonology

(40) Verb plus Clitic VV Outputs

a. deletable stem-final V plus full V

/A + e/, h + d e

/A + i/, /i + i/ —> i

/A + aJJi + aJ a

b. full V plus full V

/a + d

/a + i/

/i + i/

/a + a/

—> e

—> i

—> i

-» a

(if no h)

c. deletable stem-final V plus short V

Λ + a/, /A + a/ —> 3

d. full V plus short V

/a + a/ —> ae

Except in (40.d), the clear generalization is that the clitic V surfaces while
the verb-final V is deleted. An example of (40.a) is dative LoImpfP i-jdll-\a-s
'he goes for him' with LoImpfP /-jdllA-/. Examples of (40.b) are l-qviAkmeet
'he killed them-Fe', t-aqvAe 'she killed him', andl-rpra-\0-hi 'he killed me',
all from PerfP -3ητα-, and i-s-dlhAe 'he made him weep' from LoImpfP
-s-alha-. (40.c) occurs in imperative nazzAad ... 'sell...' (with following NP).
The aberrant case is (40.d), as in PerfP 0-osae-\dd 'he came' from -osa-.
This pattern is limited to verbs with following Centripetal clitic Asdd (and
therefore affects /a/ before a CC cluster). Since the Centripetal clitic also has
positional allomorphs without an initial V (e.g. Ad), it may be that 0-osaeAdd
is really from /osaAdd/ with no VV-Contraction at all. In either case, the
shortening to as is probably due to a special shortening rule rather than to
VV-Contraction itself. This could be taken care of formally by modifying
Presuffixal α-Shortening (§3.4.9.1). However, the phonology here is less than
transparent.
In my data (which are not complete for all dialects), the 3rd person object
clitics other than 3MaSg Ae, namely 3FeSg Aet, 3MaPl Aen, and 3FePl Aenaet,
are among the clitics that impose word-penultimate accent on an unaccented
word. In other words, they behave accentually like the corresponding
postconsonantal allomorphs, e.g. 3FeSg Atast and 3MaSg Ataen. Technically,
this means that these clitics have a true initial V that (along with the stem-final
V) is "counted" in Default Accentuation. This corresponds to a possible
historical scenario, whereby e.g. 3FeSg Aet derives from *Aast. By contrast,
2nd person object clitics like 2FePl (i)Akmaet allow antepenultimate default
accent. Therefore either these clitics lack an initial V (though forcing the