Vowel inventory and distribution 29
4. Vowel inventory and distribution
While there are occasional discrepancies in the consonantal inventory cited in
various works, there is more variation with vowels. For one, determining the
precise inventory of Madurese vowels depends on whether one considers the
phonetic level or the phonemic level; and even then there has been disagreement
in the literature. From the standpoint of phonetic realization, there is some
agreement that there are 12 vowel qualities that manifest themselves in surface
forms. These are given in the chart in the following chart:^16
front^ central^ back^high i () ɤ u (ʊ)mid ɛ ə o, ɔlow a (^)
Of these half are either allophones that can be derived through regular phono-
logical processes or constraints or auditory distinctions that may be due to other
factors. These include , ʊ, e, o, ɨ, and ɤ. According to Stevens (1968), [] and
[ʊ] result from a rule that lowers [i] and [u] in closed syllables, though this may
actually reflect vowel length (see Cohn & Lockwood 1994 and discussion be-
low). [e] and [o] are the output of a rule that raises [ɛ] and [ɔ]. These processes
are discussed further in section 4.2. [ɨ] and [] result from the Vowel Harmony
process (referred to as Vowel Tensing in Stevens 1994 and Vowel Raising in
Cohn 1993a) discussed in the next section.
4.1 Vowel harmony
Vowel harmony is perhaps the most striking feature of Madurese phonology. It
was first detailed in the work of Kiliaan (1897, 1904) and more recently in that
of Stevens (1968, 1980, 1985, and 1994).^17 Vowel harmony is contained in the
system of paired alternating vowels that characterize the vast majority of lexical
(^16) Oka et al. (1988/1989) represent 13 different vowels. No one else reports this many.
(^17) The system has also generated interest in the theoretical literature in the work of Cohn
(1993a,b), Trigo (1987, 1991), and Anderson (1991).