Vowel inventory and distribution 37
In order to account for these instances of [i] and [u] outside the conditioning
environment, they must be posited in underlying representations. It is never ne-
cessary to posit [ɨ] and [ɤ] in underlying representations of lexical items. As the
occurrence of [ɨ] and [ɤ] can always be predicted, there is no need to include
them in the phonemic inventory of the language.
Having established the vowel harmony (or raising or tensing) process and
the phonemic inventory as the six vowels /ɛ/, /ɔ/, /a/, /ə/, /i/, and /u/, it is clear
this should be reflected in the underlying representations of lexical items. Thus,
we have the following:
(33) underlying surface
/kɛpa/ [kipɤ] ‘carry’
/ɛa/ iɤ ‘remain’
/pəkkəl/ [pəkkɨl] ‘angry’
/cɔkɔ/ [cukɔ] ‘fish’
/kɔrɛ/ [kuri] ‘fry’
All of the high vowels in the surface forms are derived via the harmony rule. In
words with non-alternating vowels, the surface and underlying forms are iden-
tical. However, in order to focus discussion and make forms maximally transpa-
rent, I include the surface form of the vowels in all roots.
A final issue regarding the vowel inventory is the status of [] as a surface
vowel. In works such as Stevens (1980, 1994) and Cohn & Lockwood (1994),
eight vowels are posited as being the principle vowels in Madurese, the eight
described above as participating in the vowel alternation. However, other
sources differ on the number of primary vowels. (I use the term ‘primary vo-
wels’ here as the authors tend not to identify the inventories as being phonemic
or phonetic.) Budi et al. (1986/1987) recognize seven vowels, as do Safioedin
(1977) and Zainudin (1978). While there seems to be some conflation of [ə], [ɤ]
and [ɨ] in Budi et al., Safioden (1977), Zainudin (1978), and Pawitra (2009) po-
sit [i], [ɛ], [u], [ɔ], [a], [ə], and [ɤ] (although transcription systems vary some-
what). Moehnilabib (1979) recognizes six vowels, corresponding to the pho-
nemic inventory posited here. Notably absent in all but Stevens and Cohn &
Lockwood is any distinction between [ə] and [ɨ]. And, in fact, no practical or-
thography ever devised for Madurese has ever recognized such a distinction
(see section 8). Those that make use of symbols other than a, e, i, o, u identify
six or seven graphemes but never include a distinct symbol for [].
What, then, is the status of [ɨ]? It might be recognized for theoretical rea-
sons, i.e. to regularize the system of alternating vowels. However, absent other
evidence, ensuring paradigm uniformity for the alternating vowel system seems
an insufficient justification. There does, however, appear to be some instrumen-
tal evidence for distinguishing [ə] and [ɨ]. Cohn & Lockwood (1994) report that