542 8 Verbal derivationmasculine agentives end in deletable hi. This is really a
as found for the same verbs in the short imperfective. The presence of an
underlying stem-final V in the masculine agentive is also indicated by PI
allomorph -an (from /-sen/), which results from VV-Contraction. This analysis
shows that the surface final-syllable accent in e-m-aeCC is an underlying
penultimate accent like that in the other cases. Note that e-m-asjj 'doer,
perpetrator' from -vji> geminates the j, resulting in the same e-m-aeCC shape
seen for inputs with a CC cluster. The plurals in (569.f), e.g. "l-m-asts-an and
l-m-aejj-αη, show χ-Erasure (§3.5.3.3), so the marked accent seen in the Sg is
replaced by default antepenultimate accent.
In (569.f), e-m-aesaeww and e-m-aeraerr show resyllabification (by
Final-CC Schwa-Insertion (44)) as with the VblN, hence (in T-ka) also with
Epenthetic-Vowel Accentuation (70) and Stem-Final Gemination (71), with
the difference that the epenthetic V appears as as instead of the usual a. This
requires a different interaction between Final-CC Schwa-Insertion (44) and
Melodic Association (82) than we get in the short imperfective (including
Imprt) of the same stems: contrast agentive e-m-aesaeww 'drinker' with Imprt
3S3w (or asu) from /aeswi/, and agentive e-m-aeraerr 'reader, pupil' with Imprt
3Y3r from /aerri/. See (45) in §3.2.3.5, and §3.2.4 for analysis. The plurals, e.g.
"i-m-aesw-αη 'drinkers', are identical to those of the other agentives in (569.f)
and show no sign of resyllabification.
There are also some agentives with
has a high full V in the first syllable and low ae in the second. The type
a-m-üdaer 'living (being)' in (569.b) reflects the replacement of -vPPvC- (with
medial geminate) by a stem-variant -uPvC-. Similar replacements for -vPPvC-
verbs are seen in their productive VblN uPaC, and in their causative
derivatives with basic shape -s-uPvC-.
Several more cases with initial high V are in (569.e,h), this time with i
instead of u. These are verbs whose imperfectives are based on -i(C)CaC- or
-u(C)CaC-. In the case of 'be ahead', the agentive has a single ζ as in the
imperfectives (Imprt izar), not the geminate seen in perfectives (PerfP
-aezzar-), further strengthening the connection between the agentive and the
imperfectives. However, 'pray' in (569.c), PerfP -amud- (dialectally -aemud-)
and Imprt umad, has the α-m-aeCaC agentive.
α-m-allez 'hungry one' (569.i) is rather isolated structurally. Because ζ is
a BLC, -alliz would be another possible phonemicization. I recorded this
agentive only for T-ka, but it is attested dialectally for Niger as "amallez"
alongside "amuläz" (LTF2 206).
ά-m-idi 'friend' (569.j) is a high-frequency noun. It is also structurally
isolated, for example in its default accent. It is probably still connectable to
some extent with the verb 'accompany' (Imprt idaw, PerfP -aeddew-), cf.
abstractive t-amm-ldawa 'friendship' including the w.
Disregarding the isolated cases in (569.i-j), the following summary points
can be made about agentives from light stems: -m- or dissimilated -n- prefix,
marked penultimate accent (χ-pen), a tendency (in the few clear cases) to align