Crothers et al. 1979:325
Crothers et al. 1979:600
Saxton et al. 1983
Crothers et al. 1979:205
Crothers et al. 1979:925
Crothers et al. 1979:895
Crothers et al. 1979:800
Crothers et al. 1979:810
Crothers et al. 1979:910
103
N
OTES
*
Thanks to Adam Albright, Victoria Anderson, Katherine Crosswhite, Bruce Hayes, Sun-Ah Jun, Pat
Keating, Robert Kirchner,
Peter Ladefoged, Pam Munro, Donca Steriade, Motoko Ueyama and Jie Zhang,
and audiences at UCLA and at the 24th meeting of the Berkeley Linguistics Society for helpful discussion
of the issues presented in this paper. Any misconceptions or inaccuracies are my own responsibility.
1
It is more difficult to make a case for positing underlying voiceless vowels in any language. Dafla (Ray
1967) and Turkana (Dimmendaal 1983) appear to be the strongest cases for languages with underlying
voiceless vowels.
2
Note that /h/ is not considered here as a voiceless vowel strictly speaking, since it lacks place features
unlike “true” voiceless vowels.
3
It has been claimed (e.g. by Cho 1993 of Comanche) that some languages have both phonological and
phonetic devoicing.
4
In Ket, the short vowels are not described explicitly described as voiceless, but they are described as
overshort and virtually absent, a description which suggests that they might in fact often be voiceless.
5
In Alabama, which lacks short vowels in final position, the last half of utterance final vowels devoice.
6
The set of consonants triggering vowel devoicing in word-final position is a subset of those triggering
devoicing in phrase-final position in Tarascan.
7
In Ainu, devoicing is more likely to occur in phrase-final position than in word-final position.
8
In Woleaian, the environment for devoicing might be more accurately described as clitic group, since a
following article or deictic inhibits devoicing.
9
In Apinaye, phrase-final vowels optionally either lengthen, devoice or become creaky.
10
Vowels may devoice before silence which I interpret to mean utterance finally.
11
In Kawaiisu, likelihood of devoicing is described as gradient; devoicing is most likely in utterance
final position and less likely the smaller the domain in which the vowel is final.
12
I
n Big Valley
Shoshoni and Southern Paiute, the level of juncture which triggers devoicing is not
explicitly stated in the sources consulted.
13
In Totonac, the only utterance final short vowels occur after voiceless consonants.
14
In Malagasy, devoicing is more likely to affect word-final vowels after voiceless consonants.
15
Among the high vowels, less peripheral vowel qualities devoice over more peripheral ones, e.g. in
Ticuna and Ainu. There seems to be a slight tendency for /i/ to devoice over /u/, e.g. Mixtec, Gadsup,
Greek, Turkish, although the opposite pattern is found in Brazilian Portugese and Tunica. In Japanese,
according to Han (1962), /
¨/ is more likely to reduce than /i/, an asymmetry which Han links to the
lesser intrinsic duration of /
¨/.
16
Interestingly, in Tongan, /a/ devoices in certain environments, but mid-vowels do not.
17
Comanche also observes a restriction against two consecutive voiceless vowels; when two potential
undergoers of devoicing are adjacent, devoicing only affects the first vowel. This
is plausibly related to
the alternating stress pattern in Comanche reported by Charney (1993). Primary stress in most words
falls on the initial syllable with secondary stresses on alternating moras thereafter, but some words have
primary stress on a non-initial syllable; in such words, pretonic syllables do not devoice (Charney
1993). Cho (1993) attributes this pattern to the leftward spread of a high tone onto preceding vowels.
18
High toned syllables could alternatively be considered accented syllables in Japanese.
19
Creaky vowels, i.e. those marked with what Miller (1965) refers to as the glottal accent, also do not
devoice. Creak also blocks devoicing in Southern Paiute (Sapir 1930).
20
See Jun et al. 1997, 1998 for detailed token by token analysis of devoicing in Korean, including
discussion of results which are not compatible with an account based purely on gestural overlap.
Chia sẻ với bạn bè của bạn: