3.2 Local assimilations and syllabification rules 65verbs. Example: for the T-md informant, 'become long (or tall)' was recorded
as PerfP sajra-, Imprt sljra-t, and LoImpfP -t-lsajru-. The Imprt form here also
shows the i...a vocalism typical of the imperfective of adjectival verbs. This
verb shows up in other dialects either as an augmented verb (PerfP -assagrae-t
in A-grm, for example), or as an unaugmented verb with final stem consonant t
(PerfP sajrat- in several dialects); see §7.3.2.2. The T-md speaker appears to
have made selective, opportunistic use of the augment option to avoid having
to resyllabify the Imprt stem.
The formulation of the Final-CC Schwa-Insertion rule is tricky, since the
epenthetic V appears as a in most cases, e.g. alaw 'be spacious!', but appears
as ae in m-aesaew 'be drunk!' and e-m-aesaeww 'drinker; water source'. In three
morphological patterns we get a. One is the VblN type for derived or long
underived verbs, where the entire stem is subject to a
schwa is the expected short V: α-m-asaww 'being drunk', α-s-ilaww 'making
spacious'. A second is the set of imperfectives like (Caus) Imprt s-ilaw 'make
spacious!' and Lolmpf -t-ilaw 'is spacious', where we again assume a stem-
wide
short imperfective of type Imprt ajal 'go!' and alaw 'be spacious!', where the
epenthetic schwa breaks up a CC cluster that separates underlying
schwa is determined by the
epenthetic vowel.
This leaves mediopassive Imprt m-aesasw 'be drunk!' and e-m-aesaeww
'drinker; water source', from -vswu- 'drink'. For m-aesaew, the basic form is
arguably /m-aeswA/ instead of /m-aeswi/, since mediopassive short
imperfectives favor
source', in the agentive interpretation 'drinker' we would probably expect a
composite
t-e-m-aeksi-t-t with stem-final i. I know of one other agentive of this type,
e-m-aeraerr 'reader, pupil' from -vrru- 'read', e-m-aesaeww in the sense
'water source' is a nonagentive nominal and might have a stem-wide
melody, but agentives e-m-asvaerr and e-m-assaeww presumably have
melody like other agentives of -vCCi> stems. Given that Imprt /aeswi/ and
/aerri/ appear as asaw (or asu) and arar, while agentives /e-m-aeswi/ and
/e-m-aerri/ appear as e-m-aesaeww and e-m-aeraerr, it is necessary to explain
why the L part of the
agentives but not in the Imprt. Pending further dialectological study, it is far
from clear that Stem-Final Gemination (which is limited to T-ka nominals,
including VblN's and agentives) is correlated with this difference in melodic
domain. For now I will simply stipulate the difference as a morpho-
phonological fact (45).