jeff_l
(Jeff_L)
#1
4.1 Noun morphology 233
(208) Possible Derivations of Plurals
'banks of pond' 'eardums' basis
-dasffe ι
-daeffe
-dsffa ι
-sasle underlying (based on Sg)
-saelle gemination Γ-f and accent χ-f
-sslla PI melody
-dsff ι
— Syncope (fails to apply)
-sail Stem-Final V-Deletion
To explain why Syncope fails to apply, it is important for the gemination
formative to apply before Syncope. Otherwise the 3 vowel would be
vulnerable to Syncope in 'eardrums'.
It is odd that /a/ is subject to Stem-Final V-Deletion in these plurals. The
closest parallel is the relatively restricted underspecified stem-final vowel /A/
in some long imperfective verb forms, which deletes word-finally but shows
up as χ before a C-initial subject pronominal suffix: LoImfP -bass 'vomit', for
/-bassA-/, cf. bassae-naet 'they-FePl vomit' (§7.2.5.1, §7.3.1.3).
The derivations in (208) do look good historically, since the putative stem-
final /a/ in the plurals is actually attested in Tayert dialects of Niger. For
example, "tidaffa" (LTF2 48) corresponds to Malian t-i-daff 'banks of pond'.
Synchronically, the componential derivations in (208) are quite dubious.
There are several ad hoc ablaut formatives in addition to the regular PI ablaut
features. The final α that would have clinched the melody is nowhere to
be seen. The PI pattern t-i-C(C)aCC and variant t-i-CuC(C) with full high V
corresponds to a more diverse range of Sg shapes. These facts suggest that a
templatic analysis may be closer to the truth.
The template in this case would be -C,(C 2 )3CC- or -C,uC(C)- with full
high V, which can be integrated to some extent if υ (full V) is treated as
bimoraic, like schwa plus a C. The correctness of a templatic analysis is not
sure; note that the onset can be one or two C's, the coda is usually CC but
occasionally just C, and there is a further option of having schwa or a full high
V as the nucleus.
In the templatic analysis, we need rules for mapping segments from the
Sg onto the PI template, as follows. The leftmost C of the input is mapped onto
Q. The rightmost C (if unclustered) or CC cluster of the input is mapped onto
the postvocalic C-position of the output, undergoing gemination if necessary to
fill two C positions, especially after schwa. If there remains an unattached
medial input C, it is mapped onto C 2 , otherwise output C 2 is left vacant. These
subrules are adequate to account for the data in (206. a-e). Additional minor
subrules would have to be posited to get the various -CuC(C)- output stem
shapes in (206.f-g), as well as the idiosyncratic outputs in (206.h).
