jeff_l
(Jeff_L)
#1
140 3 Phonology
with i in the relevant position. The verb -jujju- 'load' has PerfP -ajjujja-
showing an unreduced medial u. Its LoImpfP -t-ajaejja- with melody has
medial ae, evidently shortened from /a/ (we would expect #-t-ajajja-). The
other regular verb of this shape is 'carry (baby) on back' (LoImpfP
-t-abaebba-, VblN a-bsbb). The dialectally variable verbs -Yi)(y)mi> 'sit'
(§7.3.2.9) and -juyho- 'testify' (§7.3.2.10) have full V's in the short
imperfectives in most non-T-ka dialects (length is indeterminate in the
perfectives where phonetic i could represent hy/ or /iy/, but the V is shortened
in the long imperfectives: LoImpfP -t-araeyma-, -t-ajaeyha-. The paradigm of
'load' appears to be archaic and isolated. On the other hand, I can cite no verbs
with medial α in the relevant position that do not shorten it to ae. Other verbs
with PerfP -aPPuCCa- have melody in the (short and long) imperfectives,
e.g. -ajjussa- 'go south', LoImpfP -t-ljassu-.
The long imperfectives that show Medial V-Shortening all have a full V in
the final syllable, whose length is either created or reinforced by the ablaut
feature χ-f. However, the PerfP type -ajjujja- 'load' shows that simply having
the relevant surface phonological shape ...CUCC(VC)D(C)- is not sufficient to
force shortening. There are two ways to formalize the conditions for the
shortening rule in long imperfectives. First, one might guess that the first υ
(full V) in e.g. ...CuCC(vC)u(C)- is shortened only when the final υ is
associated with the ablaut formative χ-f (the rule therefore applies only to long
imperfectives). We will see in a moment that this will not work for the short
imperfectives. Second, we could argue that shortening occurs in
...CuCC(vC)iXC)-, without regard to the formative χ-f, but is blocked when the
stem melody is composite, like in -ajjujja- 'loaded'.
To study u-Spreading in bisyllabic stems, we must consider the plural of
VblN α-CuCaC, and the Lolmpf of causative -s-uCvC-. A VblN example is
a-budar 'being rude', PI i-büdur-aen (§8.6.1.4), with u...u. However, in the
causatives, rounding and backness are transferred rather than copied:
-s-umvm- 'suck', LoImpfP -s-imum- with i...u (§8.1.5). Conceivably, the
VblN stem might be treated as trisyllabic here, including PI i-.
u-Spreading and Medial V-Shortening have a more limited effect on short
imperfectives. Since the Shlmpf has no ablaut lengthening feature, the Shlmpf
forms for verbs like 'gape' and 'be obligatory' (cf. 116, above) have a short V
in their final syllable, so they are not candidates for u-Spreading. Interestingly,
they do not undergo Medial V-Shortening either, as seen in Shlmpf -ahhussal-
'be obligatory'. Therefore, if u-Spreading and Medial V-Shortening are to
apply at all to Shlmpf forms, they must be limited to verbs that are already
V-final. As it happens, no unaugmented V-final verb has a medial u. Therefore
the only attested stems showing u-Spreading and Medial V-Shortening in the
Shlmpf are augmented V-final stems (§7.3.1.16).
Moreover, even for the augmented V-final stems, the two rules take effect
only in the portion of the Shlmpf paradigm with V-initial subject pronominal
suffixes, all of which happen to have the shape -aeC, like 3MaPl -aen. These
are precisely the forms that omit Augment -t- (in the rest of the Shlmpf
