40 3 Phonology
frequently get -taen, as in Sg ά-dwanni 'talk (noun)', PI i-dwanni-tasn. If the
noun ends in /ay/ or /aw/, we get just -asn without the (intervocalic epenthetic)
-t-, and the semivowel is audible: ...ay-aen, ...aw-aen. An example is ae-bori
'stick', PI i-boray-sen. The two stems can therefore be assigned lexical
representations /dwanni/ and /boray/, respectively. However, the Sg type
ae-bori behaves like a V-final (not C-final) stem for purposes of assigning
allomorphs of possessive pronominal suffixes. Thus ae-bori-nin 'my stick',
with postvocalic allomorph; we would expect #ae-boray-in with
postconsonantal suffix allomorph if the stem were treated as ending in a
semivowel. So again, there are data pointing in opposite directions, perhaps
lending themselves to a distinction between lexical and postlexical rules.
The type ae-bori-nin seems to be quite regular for nouns of this type. For
common nouns, the only evidence available to the native speaker that [u] is
from /aw/ or that [i] is from /iy/ is the form of the plural, and evidently this is
not powerful enough to influence the suffixal allomorphy of the singular.
However, verbal nouns are somewhat different, in that they are closely
associated in form with inflected stems. Thus the VblN 'expressing thanks',
phonetic [a-d3o'di], can be safely assigned a representation α-jüday rather than
#a-judi since the corresponding verb (PerfP -jodasy-) has an unmistakable
final diphthong. If this isn't enough, verbal nouns ending in ...CaC have a
variant with full α instead of schwa, in this case a-juday alongside a-juday.
The representation α-juday is confirmed more directly by its own PI
i-juday-aen (varying with i-jüday-aen). However, here there is some
fluctuation in the data, and plurals of the type i-jüdi-taen are attested (rarely)
along with those of the types i-jüday-aen and i-jüday-aen.
The situation may be summarized as follows. For common nouns, the
plural (ablaut or suffixal) may require a stem representation ending in a
diphthong /ay/ or /aw/, but even in this case the Sg may behave as V-final for
purposes of its own suffixal allomorphy (which arguably involves post-lexical
processes). With verbal nouns, where the evidence for a final diphthong is
much more visible to native speakers, the Sg VblN is often (but unreliably)
treated as diphthongal.
3.1.2.6 Phonological status of medial [u] and [i] before C
Most cases of medial [u] and [i] are unproblematically assigned to u and i
phonemes. However, there are some cases where I transcribe them as
diphthongs, ay or aw. The stems in question have ablaut alternations of the
type asw versus [u], or asy versus [i]. For example, the verb 'dust off has a
Shlmpf stem kaeykaey [kae'jkaej], and a PerfP stem [a'k:ikasj] that I transcribe
-akkaykasy-. Likewise Shlmpf laewlaew 'tower (above)' and PerfP [a'hulu],
the latter transcribed -allawlaw-.
For the irregular verb 'sit' (Vrym or Vrm), the most common forms are of
the perfective stem family, e.g. PerfP [-aeq:ima-] (for T-ka, the [ae] is a backed