92 3 PhonologylSg ass-ijj-aev and 3MaPl sss-ijjae-n. In all of these heavy V-final stems,
VV-Contraction converts /a-ae/ to ae (rather than e), effectively deleting the /a/,
in the lSg and 2Sg as well as 2MaPl and 3MaPl, but only after Default
Accentuation applies.
When VV-Contraction applies over a clitic boundary (40), the two V's are
both counted for purposes of Default Accentuation. For example, consider the
combination of stem-final α and the -\a- of a dative clitic (40.b). Thus
i-taw-aenna 'it is said' (passive, LoImpfP) plus 3Sg dative clitic -\a-s gives
i-taw-aenna-\0-s (attested T-md) 'it is called (named)...', with surface word-
penultimate accent.
Similarly, when 3rd person object clitics combine with a verb ending in a,
we get special clitic allomorphs and (depending on dialect) an i or e vowel
(§10.3.1). Although the special allomorphy makes the phonology less than
transparent, the i or e is treated as a contraction of two V's, both of which are
counted. This can be seen with a heavy V-final unaugmented verb like -rvbbu-
'bring up, raise (a child)'. In 3MaSg PerfPl-rrabba 'he raised, brought up (e.g.
a child)' we have default penultimate accent. If we add a 3MaPl object clitic
we get a postvocalic allomorph (e)-\n, arguably underlying Masn/, and the
result is (for T-ka) i-rrSbbe-\n 'he raised them-Ma', with surface penultimate
accent.
The contextually variable ordering of Default Accentuation and
VV-Contraction is also observable in PI suffixation in nouns. Consider the two
examples in (68), which are lifted from a much fuller data set in §4.1.2.13, see
also (39) and §3.5.3.3.
(68) Suffixal Plural of Noun
singular plural glossa. e-m-aesli l-m-aesl-αη 'voice'
b. ά-laeda i-laed-an 'fly (insect)'In (68.a), the PI has an underlying form /i-m-aesli-aen/. VV-Contraction
combines /iae/ as /a/ (39.a), and Lexical Accent Erasure (139) is triggered by
the loss of a V, giving /i-m-aesl-an/, which then undergoes Default
Accentuation to yield i-m-assl-αη with surface antepenultimate accent. By
contrast, in (68.b), Default Accentuation must apply first, giving /i-lseda-asn/,
followed by VV-Contraction (39.a), producing i-laed-αη. In the latter, my
convention is to write an acute accent, but the accent is neither lexical
(compare Sg ά-laeda) nor ablaut-induced, so it can only come from Default
Accentuation at an early stage in the derivation.
Clearly the interaction between Stem-Final i/A-deletion (29), VV-
Contraction, and Default Accentuation is complicated. It is possible to account
for the data using phonological derivations as sketched here. However, the