User:Urszag/Vowel devoicing in Japanese

Japanese vowels are sometimes phonetically voiceless. Although the difference between voiced and voiceless versions of a vowel is not phonemically contrastive, the use of voiceless vowels is often described as an obligatory feature of standard Tokyo Japanese, in that it sounds unnatural to use a voiced vowel in positions where devoicing is usual. However, the actual occurrence of devoicing is complicated by various factors.

Devoicing is primarily affected by the identity of the vowel and of the immediately surrounding segments. For the short high vowels and, it is possible to distinguish contexts where devoicing almost always occurs, contexts where it almost never occurs, and contexts where it occurs variably. In contrast, long vowels and non-high vowels show low rates of devoicing in any context.

The linguistic analysis of vowel devoicing is disputed. Among the unresolved questions are whether it should be analyzed as a phonetic process or a phonological one; whether high vowels are devoiced or outright deleted; and whether all kinds of devoicing (including both contexts where it is typical and contexts where it is atypical) are caused by one common underlying process, or whether vowel devoicing is caused by multiple distinct processes (such as a categorical phonological rule in addition to a gradient phonetic rule).

Typical devoicing environments
Devoicing mainly affects the short high (close) vowels and  in certain contexts involving adjacent voiceless consonants. Devoicing does not normally occur when these vowels are either preceded or followed by a voiced consonant or by another vowel, although occasional exceptions to this have been observed.

/i u/ between voiceless consonants
Between two voiceless consonants, high vowels are categorically, or more or less obligatorily, devoiced (apart from in certain blocking environments described below):


 * {| cellpadding="5"


 * 靴
 * 悲観
 * }
 * 悲観
 * }
 * }

Devoicing of and  between voiceless segments occurs even in careful pronunciation; it is not restricted to fast speech.

However, it is inhibited in certain contexts:

Inhibition of devoicing based on the place and manner of the surrounding consonants
In general, a high vowel between voiceless consonants is consistently devoiced when the consonant following the vowel is a stop or affricate, or when the preceding consonant is a stop and the following consonant is a fricative other than.

Devoicing seems to be inhibited before. Even though is phonemically analyzed as a voiceless consonant, it may in fact be realized with some phonetic voicing as, particularly in word-medial intervocalic position.

Devoicing is also inhibited (to a somewhat lesser extent) when a vowel is both preceded and followed by a fricative, or preceded by an affricate and followed by a fricative.

Avoidance of consecutive devoicing
When two consecutive syllables (or moras) contain high vowels between voiceless consonants, there is a tendency to avoid devoicing both vowels. The avoidance of consecutive devoicing is not absolute; Vance 2008 describes pronunciations with consecutive devoicing as common, although Fujimoto 2015 suggests that pronunciations with consecutive devoicing are less frequent than pronunciations where only the first vowel is devoiced, citing 84 cases of consecutive devoicing vs. 171 cases of devoiced-voiced sequences in word-internal sequences of devoiceable vowels from Maekawa and Kikuchi 2005's study of the CJS corpus. Tsuchida 2001 argues that when consecutive devoicing occurs, it tends to be restricted to fast speech.

When consecutive devoicing is avoided, a complex array of factors determine which vowels remain voiced and which undergo devoicing; some attempts have been made to model this, but it is sometimes described as unpredictable.


 * When two devoiceable moras occur consecutively across a word boundary, speakers prefer to devoice the vowel in the second, word-initial mora rather than the vowel in the first, word-final mora. To some extent, a parallel preference may exist when two devoiceable moras occur consecutively across a morpheme boundary in a compound word. Hirayama 2009 suggests this effect could be a consequence of a foot boundary, rather than being directly conditioned by a morpheme boundary.
 * When two devoiceable moras occur consecutively within a word, the following preferences seem to apply in order:
 * If one vowel is accented, speakers tend to prefer to devoice the other, unaccented vowel.
 * Otherwise, if one vowel is in a word-initial syllable, speakers tend to prefer to devoice it. (Devoicing has been found to occur more frequently in word-initial compared to word-medial moras.)
 * Otherwise, if one vowel is after a fricative, speakers tend to prefer to devoice it.
 * When three devoiceable moras occur consecutively within a word, devoicing usually affects the first and third while being inhibited on the second. Kawai et al. (1995)'s study of the NHK’s pronunciation dictionary found its entries showed an 84.7% rate of this devoiced-voiced-devoiced pattern for three-mora devoiceable sequences.
 * Sequences of four or more devoiceable moras are relatively rare. Fujimoto 2015, citing Kawai et al. (1995), suggests that when four devoiceable moras occur consecutively, it is preferred to devoice the first and third. However, acoustic studies have found inconsistent results.

Examples of vowels in consecutive devoicing contexts:
 * {| cellpadding="5"


 * 奇しくも
 * 菊池 寛
 * 歴史的 (Heiban)
 * 腹式呼吸
 * 焚き付ける
 * 石室
 * }
 * 歴史的 (Heiban)
 * 腹式呼吸
 * 焚き付ける
 * 石室
 * }
 * 焚き付ける
 * 石室
 * }
 * 石室
 * }
 * }


 * {| cellpadding="5"


 * 気質
 * }
 * }

/i u/ before a pause
A high vowel ( or ) may also be devoiced before a 'pause' if the preceding consonant is voiceless. A pause of this type does not typically occur within a word, so examples of devoicing in this context are essentially limited to vowels followed by a word or phrase boundary:


 * {| cellpadding="5"


 * 圧
 * }
 * }

Although devoicing between a voiceless consonant and a pause is sometimes presented as a categorical rule, devoicing in this context seems to occur with less overall consistency than devoicing between voiceless consonants.


 * Within a phrase, devoicing occurs with high frequency and consistency, regardless of speech rate, when a word-final vowel is preceded by a voiceless consonant and followed without pause (or with little pause) by a word that starts with a voiceless consonant. The presence of a longer pause after a phrase-internal word boundary seems to reduce the likelihood of devoicing before a voiceless consonant. Kilbourn-Ceron and Sonderegger 2017 suggest this is related to speech planning limitations: if a speaker is still unaware of whether the following word will start with a voiceless or voiced segment when planning how to pronounce a phrase-internal word-final vowel, this may inhibit devoicing, even if the vowel ultimately ends up being surrounded by voiceless consonants. When a sequence of two consecutive devoiceable moras occurs across a word boundary, the vowel in the word-initial mora is more likely to be devoiced than the preceding word-final vowel.
 * Compared to phrase-internal word-final vowels, vowels at the end of a phrase are less consistently devoiced before words starting with a voiceless consonant. A vowel between voiceless consonants is devoiced most consistently within a word, next at the end of a word in the middle of a phrase, next at the end of an accentual phrase, and least consistently at the end of an intonation phrase (Kilbourn-Ceron and Sonderegger's model estimates that devoicing rates before an intonation phrase boundary range between 56% and 93%). Phrase-final vowels are devoiced more frequently in faster speech rates. Unlike in phrase-internal position, devoicing of phrase-final vowels is positively associated with the presence of a longer pause after the word boundary: the rate of devoicing may be nearly categorical when a phrase-final vowel is followed by a pause longer than 463 milliseconds.
 * Final is frequently devoiced in the common sentence-ending copula です and polite suffix ます.
 * Phrase-final vowels are not devoiced when the phrase carries the rising intonation associated with an interrogative sentence, as in the question 行きます?.
 * Labrune states that devoicing between a voiceless consonant and a pause mainly affects unaccented vowels. Per Fujimoto 2015, not all studies show consistent devoicing of final and  between a voiceless consonant and a pause, even when unaccented.

/i u/ next to a voiced consonant
A high vowel after a voiceless consonant may occasionally be devoiced even when the following segment is voiced. Devoicing in this context seems to be most frequent when the following consonant is a nasal or an approximant. Per Vance 2008, high vowels are not devoiced next to a voiced segment in careful pronunciation.

Some studies have found rare examples of voiceless vowels after voiced consonants.

/a o e/ between voiceless consonants
The non-high vowels are sometimes devoiced, usually between voiceless segments; devoicing in this context is infrequent, optional, varies between speakers, and can be affected by speech rate. Different authors give slightly different descriptions of the conditions where non-high vowels are prone to be devoiced.


 * Non-high vowels are reportedly more likely to be devoiced when they occur in the first of two identical moras, as in kakashi.
 * Labrune 2012 describes non-high-vowel devoicing as being restricted to cases where the vowel comes between voiceless consonants and the same vowel occurs in the following mora (without specifying whether the surrounding consonants are the same or different).
 * The NHK Accent Dictionary (1985) described devoicing of as being more likely to occur when the moras  or  occurred twice in sequence or when  or  were followed by a mora with the same vowel.
 * They may be more likely to be devoiced after a fricative than after a plosive.
 * They may be more likely to be devoiced when they are unaccented.
 * Some descriptions state that are more prone to being devoiced than /e/, but it is not clear that this is true.

Examples of voiceless non-high vowels:
 * {| cellpadding="5"


 * 心
 * 墓
 * }
 * 墓
 * }
 * }

However, Tsuchida gives haka as an example of a word where the vowel in the first mora is not devoiced.

It is disputed whether devoicing of high and non-high vowels is caused by the same process.
 * Hirayama 2009 argues for a unified explanation of devoicing for high and non-high vowels, with high vowels simply being more susceptible to devoicing as a result of their inherently shorter duration.
 * Martin, Utsugi & Mazuka 2014 argue that high vowel devoicing is phonological and intentional, while low vowel devoicing occurs unintentionally as the result of gestural overlap.

Phonetics of devoiced vowels
Different sources give somewhat different descriptions of the phonetic realization of devoiced vowels.


 * A devoiced vowel may sound like a period of turbulent noise showing the acoustic characteristics of a voiceless fricative: for example, the devoiced of kitai sounds like the voiceless palatal fricative.
 * Sometimes there is no clear acoustic boundary between the sound of a devoiced vowel and the sound of the preceding voiceless consonant phoneme. For example, although the word /suta↓iru/ is phonemically analyzed as containing a devoiced /u/ in its first mora, it can be acoustically analyzed as starting with a fricative [s] that is sustained up until the following [t], with no third sound intervening between these two consonant sounds.

Recent research has argued that "vowel deletion" more accurately describes the phenomena. Based on the absence in some circumstances of a segment displaying the acoustic properties of a vowel, it has been proposed that 'devoiced' vowels are in fact deleted (either phonetically, or phonemically). However, it has been argued in response that other phenomena show at least the underlying presence of a vowel phoneme:


 * Even when the vowel of a CV sequence is devoiced and appears to be deleted, the pronunciation of the preceding consonant phoneme shows coarticulatory effects.
 * Prosodically, vowel devoicing does not affect the mora count of a word.
 * When a vowel is devoiced between two identical voiceless fricatives, the result is typically not pronounced as a single long fricative: usually, two acoustically distinct fricative segments are produced, although it may be difficult to describe the acoustic characteristics of the sound that separates them. In this context, alternative pronunciations involving a voiced vowel are more common than they are between other voiceless sounds. Vowel between two identical voiceless fricatives may have either a weak voiceless approximant release or a revoiced vowel depending on the rate of speech and individual speech habits.
 * {| cellpadding="5"


 * 日進橋
 * vs.
 * or
 * 西新橋
 * 決済
 * vs.
 * or
 * 消す際
 * }
 * or
 * 消す際
 * }

Sociolinguistic and stylistic aspects of vowel devoicing
Per Shaw & Kawahara 2021, Imai 2004 describes devoicing as a marker of prestigious speech in Tokyo Japanese.

In typical devoicing environments, high vowels can be devoiced at all speech rates and levels of formality. Different studies have found different effects of speech style on the frequency of devoicing.

Martin, Utsugi, & Mazuka 2014 found that in comparison to spontaneous adult-directed speech, carefully read speech shows a higher rate of high vowel devoicing and a lower rate of non-high-vowel devoicing, whereas infant-directed speech shows a lower rate of high vowel devoicing and a higher rate of non-high-vowel devoicing. Likwise, Hasegawa 1979 described devoicing as less likely to occur in casual speech. In contrast, other studies have found devoicing to be more frequent in spontaneous or casual speech than in controlled or formal speech.

The occurrence of devoicing varies between different regional varieties of Japanese. The use of voiced vowels in contexts that would typically trigger devoicing has been described as a characteristic of some varieties spoken in Western Japan or the Ryukyu Islands. Some nonstandard varieties of Japanese can be recognized by their hyper-devoicing.

Different studies have found different effects of gender on vowel devoicing.

It is regarded as effeminate to prolong a terminal vowel, particularly the terminal as in あります.