48 3 Phonology
C-initial perfectives, before which t- fails to appear, but these stems avoid
3MaSg i- (and often 1P1 n-) as well, so in this case the absence of t- is entailed
by a more general avoidance of subject prefixes.
(35) Prefixal t-Deletion
A pronominal subject prefix of the form /t-/ is deleted before a
C-initial verb stem.The rule does not apply to nominal morphology, where Fe prefix t- is
never deleted. In the rare cases (involving loanwords) where Fe t- directly
precedes a C-initial stem that lacks a vocalic prefix, Schwa-Epenthesis applies
and we get ta- (§3.2.5). In several dialects, though not in T-ka, the FePI prefix
combination t-i- reduces (by Prefix Reduction) to t-0- before a stem beginning
in CV..., e.g. PI t-l-kabr-en 'sparrows', but reduced form (in a PP) daer
^t-O-kabr-en 'in the sparrows'. In this combination, the t- is not deletable.
(T-ka reduces -i- to schwa here: daer "t-a-kabr-en.)
Whether Prefixal t-Deletion also applies to the -t- prefix marking the long
imperfective stems in many verb classes is a tricky question. In those cases
where it appears audibly, this -t- is always followed by a V, so there is no
question of deletion. However, all cases (except one) where -t- is absent from a
long imperfective stem involve C-initial stems like LoImpfP -bdddasd- 'get
up' and causative -s-arad- 'wash'. Although I do not favor this analysis, one
might argue that -t- is underlyingly present here but is always deleted by
Prefixal t-Deletion.
The one case where -t- is (or rather may be) absent from a long
imperfective stem is in the class of verbs of basic shape -vPvC-. Here the
LoImpfP varies between -aPPaC- and -t-aPPaC-, e.g. -awwat- and -t-awwat-
'hit'. We therefore have only equivocal evidence as to whether -t- is part of all
long imperfective stems. The alternative analysis is that -t- is in
complementary distribution with F-c2 (i.e. gemination of C 2 ); see §3.4.2.1.
In the type -aPPaC- varying with -t-aPPdC-, if a t- subject prefix is present
we always get e.g. phonetic ftaw:a't], which can be parsed morphologically as
either t-awwdt (with 3FeSg t-) or as 0-t-awwat from /t-t-awwat/.
3.2.2 Longer-distance consonantal interactions
3.2.2.1 Consonantal metathesis
There are no truly productive metathesis rules. The cases discussed here
involve the consonantal sequences (often with intervening V, sometimes also
with an intervening C) in (36).