jeff_l
(Jeff_L)
#1
3.5 Syntactically controlled phonological processes 157
The reason why the 3FeSg subject forms are singled out for this special
treatment is not hard to see. For a Lolmpf like -tatt- 'eat', the 3FeSg subject
form is the only form that (after deletion of 3FeSgS prefix t-) has neither an
audible (and generally syllabic) subject prefix that would obviate Rightward
Accent Shift, nor a syllabic subject suffix that (after a monosyllabic stem)
could host a rightward-shifted accent. This applies not only to -tdtt- but to all
LoImpfP stems, since they are overwhelmingly C-initial. Admittedly, there is
one verb type, namely -vPvC-, that has two alternative LoImpfP stem shapes,
C-initial -t-aPPdC- and V-initial-aPPdC- (§7.3.1.1). In theory, the V-initial
variant would allow 3FeSgS prefix t- to be audible. However, given that
-t-aPPciC- and -aPPaC- are in free variation, and that 3FeSgS t- is zeroed
before a C, when we hear a verb form [taPiaC] there is no way to tell whether
this represents t-aPPdC- with audible 3FeSgS t- plus stem -aPPaC-, or
0-t-3PPdC- with zeroed 3FeSgS prefix plus stem -t-aPPdC-.
Since LoImpfP stems like -baddaed- 'stand up' retain the ablaut-induced
accent (from formative χ-pcl) even while erasing the ablaut-induced
lengthening (formative χ-pcl), producing -baeddaed- or with Rightward
Accent Shift -baeddaed- (e.g. in 3MaPl basddaed-asn), this morphological form
(i.e. LoImpfP in definite relative) is distinct from the Prohibitive stem, which
completely lacks χ-pcl as well as χ-pcl, hence -baeddaed- (e.g. waer basddsed
'don't stand up!', arguably with deleted It-/ prefix).
The cases we have considered represent the only verb stems where
Rightward Accent Shift runs afoul of a rule deleting stem V's. This suggests a
derivation for unaccented 0-tastt in (134.a) where Rightward Accent Shift in
fact does shift accent onto the stem-final /A/, whereupon Stem-Final i/A-
Deletion (29) zeroes the /A/, and the ablaut-induced accent on the zeroed V
disappears. Technically, we could think of this as a delinking of the accent,
which ends up with no vowel to attach to.
However, there is another construction involving accent erasure that must
also be considered before any rules are formulated. Like the process affecting
3FeSg subject definite relatives, this one involves erasure of a marked accent
when a stem-final V disappears. However, this time we are dealing with nouns
rather than verbs, the loss of the V is due to VV-Contraction (39.b) rather
than Stem-Final i/A-Deletion (29), and there is no phonological reason why
the contracted vowel could not host the accent. Moreover, there is no reason to
think that Rightward Accent Shift is involved.
The relevant cases here are agentives of the type e-m-asqr 'killer', PI
"i-m-aeqr-an (§8.8.1), including causative agentives of the type e-m-aes-aeqq
'cook', PI i-m-ses-aeqq-an (§8.8.4), and VblN's of the type α-fsyk 'being
searched', PI Ί-fayk-an (§8.6.1.4). In these forms, there is a marked accent in
the Sg that disappears in the PI. These agentives and VblN's are based on
V-final stems, here -νηγυ- 'kill', -νηηυ- 'be cooked', and -fvyku- 'be
searched'. In the Sg forms, we can assume that Stem-Final i/A-Deletion (29)
has deleted a final V, so the surface final-syllable accent can be derived from a
(marked) penultimate accent (i.e. with χ-pen), e.g. underlying agentive
