93
THE PHONETICS AND PHONOLOGY OF NON-MODAL
VOWELS: A CROSS-LINGUISTIC PERSPECTIVE
M
ATTHEW
G
ORDON
University of California, Los Angeles
This paper provides phonetic explanations for a number of asymmetries in the
distribution of non-modal voiced and modal voiced vowels found in a typology
of over 50 languages with non-modal vowels. The distribution of non-modal
vowels cross-linguistically is argued to result from the conflicting demands of
perceptual salience and articulatory ease; this conflict can be formally modeled in
an constraint-based grammar.
*
1. I
NTRODUCTION
. Some languages of the world have vowels characterized by non-
modal phonation, e.g. breathy voiced vowels, voiceless vowels, or creaky
(laryngealized) vowels. Depending on the language and on the phonation type,
these non-modal vowels may either contrast with or be allophonic variants of modal
voiced vowels. For example, creaky vowels are phonemic in Kedang (Samely 1991)
and Jalapa Mazatec (Kirk et al. 1993), but occur allophonically in the vicinity of
glottalized consonants in many languages, e.g. Georgian and Tzeltal (cf. Crothers et
al. 1979). Breathy voiced vowels are phonemic in Gujarati (Fischer-Jørgensen
1967) but occur allophonically in the vicinity of /
h/ in many languages. Similarly,
voiceless vowels contrast with voiced vowels on the surface in Turkana
(Dimmendaal 1983), but occur allophonically in Japanese (Han 1961).
Non-modal vowels have a quite different distribution from modal vowels. First,
they are quite rare cross-linguistically, both as phonemic segments which contrast
with modal vowels, and as non-contrastive allophones of modal voiced vowels. For
example, Maddieson’s (1984) survey of 317 languages includes only two with
contrastive laryngealized/creaky voice (Sedang and Southern Nambiquara), two with
contrastive voiceless vowels (Ik and Dafla), and one with phonemic breathy voiced
vowels (Tamang). Another characteristic property of non-modal vowels which
differentiates them from modal vowels is their limited distribution. For example,
voiceless vowels are often limited to word-final position, and creaky vowels tend to
occur adjacent to glottalized consonants. In other languages, non-modal vowels are
the synchronic manifestations of other types of contrasts, e.g. segmental, tonal, or
durational ones.
Given the limited distribution of non-modal vowels relative to modal vowels
there are a couple of basic questions which come to mind. First, why do non-modal
vowels typically play a limited role in the phonology of most languages? Second, is
the distribution of non-modal vowels predictable on phonetic grounds? This goal of
this paper is to provide answers to these questions and to formalize these answers in
an Optimality Theoretic grammar.
2. T
HE RARITY OF NON
-
MODAL VOWELS CROSS
-
LINGUISTICALLY
. I conjecture
that the rarity of contrastive non-modal vowels has a perceptual basis; non-modal
vowels are perceptually less robust than modal vowels and are therefore eschewed
by many languages. It has been shown by Silverman (1995, 1998) that, non-modal
phonation reduces the ability of vowels to manifest tonal contrasts in a salient
manner. Given this fact, it is thus not surprising that many tone languages (e.g.
Jalapa Mazatec) restrict overlap between tonal and phonation contrasts.
However, beyond the inherent incompatibility between non-modal voicing and
tone discussed by Silverman, there is reason to believe that non-modal vowels share
properties which make them inherently less salient than modal vowels, even in the
94
absence of tonal contrasts. This reduced salience can inhibit the recovery of
contrastive place information in the vowel. Let us now consider the acoustic
properties which make non-modal vowels less salient than modal vowels.
First, non-modal vowels are characterized by less overall acoustic intensity than
modal vowels as shown for different non-modal phonation types, e.g. breathy
vowels in Kui and Chong (Thongkum 1987), creaky vowels in Chong (Thongkum
1987) and voiceless and creaky vowels in Hupa (Gordon 1998). Decreasing the
acoustic intensity results in a decrease in loudness, the auditory correlate of
intensity; it thus follows that non-modal vowels are less salient than modal vowels.
Furthermore, non-modal voicing often alters the spectral properties of vowels,
including formant structure, as demonstrated instrumentally for Kedang (Samely
1991) and Chong (Thongkum 1987), and also qualitatively evident in many
languages, e.g. those in which vowels in the vicinity of glottal stop and /h/ have
noticeably different qualities than vowels in other environments not associated with
non-modal phonation (cf. Blankenship 1997 for extensive discussion of the spectral
properties of non-modal vowels). The perturbation of formant structure potentially
makes recovery of vowel quality contrasts more difficult. In summary, given the
reduced intensity of non-modal phonation and its influence on vowel quality, it
would thus not be surprising that many languages design their phonologies to limit
the distribution and the role of non-modal vowels.
In addition to the general paucity of contrastive non-modal vowels cross-
linguistically, other language specific facts suggest that phonologies avoid non-
modal vowels for perceptual reasons. First, breathy voiced vowels in Kedang
(Samely 1991), and both breathy and creaky vowels in Jalapa Mazatec (Kirk et al.
1993) are phonetically much longer (up to 50% longer) than their modal voiced
counterparts. An interesting aspect of Jalapa Mazatec is that non-modal voicing
does not persist throughout the entire duration of phonemic breathy and glottalized
vowels. Rather, non-modal voicing occurs principally on the first half of non-modal
vowels; the second half of non-modal vowels is phonetically characterized by modal
voicing. Acoustic measurements of fundamental frequency and formants suggest
that non-modal voicing is also largely confined to the first half of the vowel in
Kedang. Crucially, Kedang is not a tone language; thus, the realization of non-
modal voicing cannot be attributed to the presence of tonal contrasts. Rather,