What are the actual underlying phonemes of Japanese?

A forum for discussing linguistics or just languages in general.
Post Reply
User avatar
Isfendil
greek
greek
Posts: 668
Joined: 19 Feb 2016 03:47

What are the actual underlying phonemes of Japanese?

Post by Isfendil »

Ever since I started linguistics, I learned that Romanji has been lying to me (given that I don't watch anime, it took me that long to find out) and that so many sounds in Japanese are actually allophones. Wikipedia takes far too long to get to the point I'm trying to reach, so... what exactly is Japanese's underlying, unallophoned consonant inventory? Let's pretend that none of the allophones exist, for instance. What's left?
Sumelic
greek
greek
Posts: 566
Joined: 18 Jun 2013 23:01

Re: What are the actual underlying phonemes of Japanese?

Post by Sumelic »

Figuring out the "actual" phonemes of a language isn't straightforward. For example, some English people pronounce "Bach" with [x] (or [χ]). Does that make /x/ an English phoneme? Is the English word "hue" composed of three phonemes /h/+/j/+/uː/, or two phonemes /h/+/i͡u/, or two phonemes /ç/+/uː/? Is /ŋ/ a phoneme, or is it underlyingly /n/ or /ng/, with the apparent /ŋ/~/ng/ contrast in words like "singer" vs. "finger" only arising for morphological reasons?

I'm not that familiar with Japanese, so someone like clawgrip could give a better explanation, but my understanding is as follows. For most modern native Japanese speakers:
  • there is still a strong neutralization of the contrast between and [ʃ] before /i/, with most Japanese speakers having trouble producing or perceiving [si] as a distinct sound from [ʃi]. The two sounds contrast before /a o u/ in many native words. Before /e/, generally only occurs in native words, although apparently the monosyllabic exclamation /ʃee/ can occur in colloquial speech (I wonder if this arose from monopthongization of /ai/; e.g. [ʃai] > [ʃee]). Speakers generally don't neutralize [se] and [ʃe] in modern loanwords. Source:

    Code: Select all

    https://books.google.com/books?id=8vFeCAAAQBAJ&pg=PA139&lpg=PA139&dq=Japanese+s+ʃ+contrast+e&source=bl&ots=6wGTESZoQD&sig=da9XQcKRCQr-vQg9mYiBJllx_tc&hl=en&sa=X&ved=0ahUKEwiJrK3wmoTRAhVJ4oMKHUVbDF4Q6AEIHDAA#v=onepage&q=Japanese%20s%20ʃ%20contrast%20e&f=false
  • [t] and [tʃ] are clearly contrastive before /a o/ and pretty easily contrasted before /e i/. In native words, [tʃe] and [ti] don't exist, and [tʃi] occurs in verb inflection when a morpheme ending in /t/ is followed by a morpheme starting with /i/. Before /u/, [t] doesn't exist in native words, [ts] being used instead. This clearly contrasts with [tʃ] and it seems Japanese speakers also don't have much difficulty pronouncing [tu], although it seems it shows up less often than [ti] in loanwords (due I think to the Japanese system for adapting English vowels). [ts] only occurs before /u/ in native words; I'm not really familiar with the situation for [ts] in other contexts. It definitely wouldn't be easy for a Japanese speaker to contrast it with [tʃ] before /i/, but it seems Japanese speakers don't have much trouble pronouncing it before /a e o/, although since English doesn't have /ts/ there probably aren't so many loanwords with [tsa tse tso].
  • [d], [dʒ] and [dz~z] are all potentially contrastive before /a e o u/. The contrast between [dz~z] and [dʒ] is probably strongly neutralized before /i/, just as for vs. [ʃ]. The sequences [dʒe], [di], [du] only show up in loanwords, however, and [dʒi] and [dzu~zu] occur in verb paradigms for d + i and d + u.
  • [h] and [ɸ] are contrastive before /a e i o/, but not before /u/. The sound [ɸ] occurs before /a e i o/ only in loanwords. Either [h] or [ɸ] or something in between may be used before /u/. As far as I know, there aren't any paradigmatic alternations between [h] and [ɸ].


I think there's a strong argument for transcribing Japanese with the phonemes

/p b t d k g (Q)/
/ɸ ts z s tʃ dʒ ʃ h/
/m n (N)/
/r w j/

assuming you're just giving a relatively shallow phonemic analysis and not the kind of one that tries to include all morphological processes.

Actually, the above table is a bit inconsistent in a way that might reflect my biases as an English speaker. It would probably be more consistent to either remove /tʃ dʒ ʃ/ (as these can be analyzed as /tsj/ or /ts/, /zj/ or /z/, or /sj/ or /s/), or to add in the rest of the palatal/palatalized consonants such as /ç/, /bʲ/, /mʲ/, /kʲ~c/ etc. But a di-phonemic analysis of [tʃ] as /tsj/ seems a bit problematic to me since it actually occurs in a wider range of contexts than than [ts], which is the opposite of what would be expected for a cluster. I guess it would also work to analyze it as /ts/ before /i/, and /tj/ in other contexts, but that's also a bit inconsistent.

Maybe it does make sense to treat /tʃ dʒ ʃ/ as special after all since it seems that the contrast between [ʃe] and [se], [dʒe] and [de] or [(d)ze], and [tʃe] and [te] or [tse] is fairly well established while there is no contrast between things like [çe] and [he] or [kʲe] and [ke].
Last edited by Sumelic on 21 Dec 2016 04:42, edited 5 times in total.
User avatar
sangi39
moderator
moderator
Posts: 3024
Joined: 12 Aug 2010 01:53
Location: North Yorkshire, UK

Re: What are the actual underlying phonemes of Japanese?

Post by sangi39 »

From what I can tell, excluding distinctions that have come into the language from loans from other languages like English, the underlying phonemes of Japanese are:

/p b t d k g/
/m n/
/s z h/
/ɺ/
/j ɰᵝ (also transcribed /w/)/

/i ɯᵝ (also transcribed /u/)/
/e o/
/a/

Some people consider the "moraic ɴ" either /n/ in coda position or as a distinct phoneme.

In native vocabulary, the main allophonic variation affects /ti tj tu di dj du si sj zi zj/ which appear as [tɕi tɕ tsu dʑi dʑ dzu ɕi ɕ ʑi ʑ] although [dzu] often appears as [zu] and [ʑi] as [dʑi], neutralising the distinction between /di zi/ (and similarly /dj zj/) and /du zu/for a number of speakers.

/hu/ appears as [ɸu] and /hj/ appears as [ç].

And that, as far as I know, covers the "major" allophonic processes in Japanese that lead to tables like the one found on Wikipedia.

As Sumelic points out, though, the influx of loan words into Japanese that haven't been "fully nativised", i.e. they haven't been stripped down completely to match Japanese phonology and then undergone allophonic changes where appropriate, have lead to distinctions such as [ti] (non-native) vs. [tɕi] (native) and [ha] (native) vs. [fa] (non-native) becoming increasingly more common, so much so that some analyses of Japanese phonology will present /t/ and /tɕ/ as distinct phonemes.



Most Romaji you'll see are variants of Hepburn, which takes this allophony into account in order to more accurately represent pronunciation. Nihon-Shiki and Kunrei-Shiki (although the latter less so than the former), on the other hand, more strictly follow the syllabary, so Hepburn has ji (closer to the pronunciation), NS has di (strictly following d+i) and KS has zi (indicating the neutralisation of /d/ and /z/ in this particular syllable) for Hiragana ぢ.

Hepburn Romaji is strictly speaking "lying", it's just telling you the truth from a certain point of view [:P]
You can tell the same lie a thousand times,
But it never gets any more true,
So close your eyes once more and once more believe
That they all still believe in you.
Just one time.
GrandPiano
mayan
mayan
Posts: 2080
Joined: 11 Jan 2015 23:22
Location: USA

Re: What are the actual underlying phonemes of Japanese?

Post by GrandPiano »

Isfendil wrote:Ever since I started linguistics, I learned that Romanji has been lying to me
Small correction: romaji, not romanji

Sangi's analysis seems pretty thorough (and I agree with it more than Sumelic's if recent loanwords are ignored), although they left out vowel length. The difference between [a] and [aː] is usually regarded as one of phonemic vowel length, although some argue that [aː] is really /a.a/ phonemically.

Also, not exactly a phoneme, but pitch accent is a phonemic feature in Japanese rarely shown in romaji.
Sumelic wrote:[ts] only occurs before /u/ in native words; I'm not really familiar with the situation for [ts] in other contexts. It definitely wouldn't be easy for a Japanese speaker to contrast it with [tʃ] before /i/, but it seems Japanese speakers don't have much trouble pronouncing it before /a e o/, although since English doesn't have /ts/ there probably aren't so many loanwords with [tsa tse tso].
[ts] is usually regarded as an allophone of /t/ before /u/. [tɕ] (not [tʃ]) is usually analyzed as an allophone of /t/ before /i/ and an allophonic realization of /tj/ before other vowels. /ts/ as a phoneme only exists in rare loanwords such as ツァーリ tsāri "czar".
Sumelic wrote:
  • Either [h] or [ɸ] or something in between may be used before /u/.
I think /hu/ is always pronounced [ɸu], never [hu] (which is why it's usually romanized as fu).
Sumelic
greek
greek
Posts: 566
Joined: 18 Jun 2013 23:01

Re: What are the actual underlying phonemes of Japanese?

Post by Sumelic »

GrandPiano wrote:
Sumelic wrote:[ts] only occurs before /u/ in native words; I'm not really familiar with the situation for [ts] in other contexts. It definitely wouldn't be easy for a Japanese speaker to contrast it with [tʃ] before /i/, but it seems Japanese speakers don't have much trouble pronouncing it before /a e o/, although since English doesn't have /ts/ there probably aren't so many loanwords with [tsa tse tso].
[ts] is usually regarded as an allophone of /t/ before /u/. [tɕ] (not [tʃ]) is usually analyzed as an allophone of /t/ before /i/ and an allophonic realization of /tj/ before other vowels. /ts/ as a phoneme only exists in rare loanwords such as ツァーリ tsāri "czar".
There is no difference between [tɕ] and [tʃ] in the context of Japanese. It's just a matter of broad vs. narrow transcription, like = [ɯ] = [ɯβ]; you can see the book I cite also uses [tʃ] and I've seen it used in various papers on Japanese phonology. I read a bit more of that book, and it says [tu] actually is generally less stable than /ti/ in loanwords, and fairly often is substituted with [tsu] even today. I think it's kind of interesting that there's this apparent asymmetry between the behavior of /ti/ and /tu/ in loanwords.
GrandPiano wrote:
Sumelic wrote:
  • Either [h] or [ɸ] or something in between may be used before /u/.

I think /hu/ is always pronounced [ɸu], never [hu] (which is why it's usually romanized as fu).

There seems to be a difference for many speakers in the level of friction between /ɸ/ followed by non-/u/ vowels (in loanwords) and /ɸu/.
Perhaps the most significant sound which occurred was the blend between [ɸ] and [h]. The contrast between this blended sound and the voiceless bilabial fricative evidently was caused by a lack of lip movement in the utterance of [ɸ/h] versus lip movement in the utterance of [ɸ]. This finding supports Watanabe (2009) who mentions the occurrence of lip movement in innovative /fV/ sequences where he believes [ɸ] occurs, as well as the lack of lip movement in conservative /hu/ where he believes a sound closer to [h] occurs. In fact, the blended sound seems to match well with Watanabe’s (2009) description of the pronunciation of /hu/, as I believe this blended consonant sounds more like an [h] than it does a [ɸ]. In fact, we may consider this sound to be transcribed as [hw] in order to represent the extra labialization causing extra frication in the pronunciation of this sound.

"An acoustic study of the Japanese voiceless bilabial fricative", Scott Ruddell (pp. 13-14)
User avatar
sangi39
moderator
moderator
Posts: 3024
Joined: 12 Aug 2010 01:53
Location: North Yorkshire, UK

Re: What are the actual underlying phonemes of Japanese?

Post by sangi39 »

GrandPiano wrote:
Isfendil wrote:Ever since I started linguistics, I learned that Romanji has been lying to me
Small correction: romaji, not romanji

Sangi's analysis seems pretty thorough (and I agree with it more than Sumelic's if recent loanwords are ignored), although they left out vowel length. The difference between [a] and [aː] is usually regarded as one of phonemic vowel length, although some argue that [aː] is really /a.a/ phonemically.
Huh, I didn't know that. Is it just [a:] that's analysed that way, or was it just an example and some people also analyse [u:], [i:], etc. as phonemically long vowels?


GrandPiano wrote:Also, not exactly a phoneme, but pitch accent is a phonemic feature in Japanese rarely shown in romaji.
Oh god I'd love to have a romanisation out there that depicted pitch accent but was also in widespread use.

Actually, is there a reason pitch accent isn't usually transcribed? I seem to recall reading that there aren't a great many otherwise homophonous words are distinguished by pitch accent, or at least few enough that with proficient knowledge of Japanese they can be distinguished by context (as if English /ju:/, whether referring to the sheep or the tree, were always written "yu").
You can tell the same lie a thousand times,
But it never gets any more true,
So close your eyes once more and once more believe
That they all still believe in you.
Just one time.
Squall
greek
greek
Posts: 526
Joined: 28 Nov 2013 14:47

Re: What are the actual underlying phonemes of Japanese?

Post by Squall »

The Japanese pure phonemes are represented in Hiragana.
You can learn it here.

I think it is weird because it doesn't have ɕe ʨe ʥe ɸa.
English is not my native language. Sorry for any mistakes or lack of knowledge when I discuss this language.
:bra: :mrgreen: | :uk: [:D] | :esp: [:)] | :epo: [:|] | :lat: [:S] | :jpn: [:'(]
GrandPiano
mayan
mayan
Posts: 2080
Joined: 11 Jan 2015 23:22
Location: USA

Re: What are the actual underlying phonemes of Japanese?

Post by GrandPiano »

Sumelic wrote:
GrandPiano wrote:
Sumelic wrote:[ts] only occurs before /u/ in native words; I'm not really familiar with the situation for [ts] in other contexts. It definitely wouldn't be easy for a Japanese speaker to contrast it with [tʃ] before /i/, but it seems Japanese speakers don't have much trouble pronouncing it before /a e o/, although since English doesn't have /ts/ there probably aren't so many loanwords with [tsa tse tso].
[ts] is usually regarded as an allophone of /t/ before /u/. [tɕ] (not [tʃ]) is usually analyzed as an allophone of /t/ before /i/ and an allophonic realization of /tj/ before other vowels. /ts/ as a phoneme only exists in rare loanwords such as ツァーリ tsāri "czar".
There is no difference between [tɕ] and [tʃ] in the context of Japanese. It's just a matter of broad vs. narrow transcription, like = [ɯ] = [ɯβ]; you can see the book I cite also uses [tʃ] and I've seen it used in various papers on Japanese phonology.

Fair enough. I'm used to seeing [tɕ], which is more phonologically accurate, but I guess [tʃ] works too in a broad transcription (although is significantly more convenient to type than [ɯ] or [ɯβ] while [tɕ] and [tʃ] are equally convenient to type). My main point was that [tɕ], if not its own phoneme, is phonemically /tj/, not /tsj/.

Sumelic wrote:
GrandPiano wrote:
Sumelic wrote:
  • Either [h] or [ɸ] or something in between may be used before /u/.

I think /hu/ is always pronounced [ɸu], never [hu] (which is why it's usually romanized as fu).

There seems to be a difference for many speakers in the level of friction between /ɸ/ followed by non-/u/ vowels (in loanwords) and /ɸu/.
Perhaps the most significant sound which occurred was the blend between [ɸ] and [h]. The contrast between this blended sound and the voiceless bilabial fricative evidently was caused by a lack of lip movement in the utterance of [ɸ/h] versus lip movement in the utterance of [ɸ]. This finding supports Watanabe (2009) who mentions the occurrence of lip movement in innovative /fV/ sequences where he believes [ɸ] occurs, as well as the lack of lip movement in conservative /hu/ where he believes a sound closer to [h] occurs. In fact, the blended sound seems to match well with Watanabe’s (2009) description of the pronunciation of /hu/, as I believe this blended consonant sounds more like an [h] than it does a [ɸ]. In fact, we may consider this sound to be transcribed as [hw] in order to represent the extra labialization causing extra frication in the pronunciation of this sound.

"An acoustic study of the Japanese voiceless bilabial fricative", Scott Ruddell (pp. 13-14)

Interesting. It definitely makes sense that there would be more frication in environments where [ɸ] contrasts with [h]. I still think /hu/ is very rarely if ever actually pronounced [hu], even if the actual pronunciation is closer to [hu] than [ɸu].

sangi39 wrote:Huh, I didn't know that. Is it just [a:] that's analysed that way, or was it just an example and some people also analyse [u:], [i:], etc. as phonemically long vowels?]

It applies to all long vowels; I was just using [aː] as an example. Interestingly, the Japanese writing system does it both ways: in hiragana, vowel length is usually indicated by doubling the vowel (あ a, ああ ā, い i, いい ī), which suggests that it's really /V.V/, while in katakana, a separate character is used (ア a, アー ā, イ i, イー ī), which suggests that it's really /Vː/.

sangi39 wrote:Actually, is there a reason pitch accent isn't usually transcribed? I seem to recall reading that there aren't a great many otherwise homophonous words are distinguished by pitch accent, or at least few enough that with proficient knowledge of Japanese they can be distinguished by context (as if English /ju:/, whether referring to the sheep or the tree, were always written "yu").

That's pretty much the reason why. Words distinguished by pitch accent alone can usually be easily distinguished through context if pitch accent is ignored, so most people who are aware of it don't consider it worth teaching in class or indicating in romaji.

Squall wrote:I think it is weird because it doesn't have ɕe ʨe ʥe ɸa.

That's because those combinations of phones don't exist outside of loanwords and colloquialisms. The table in the article you linked only lists morae occuring in native and Sino-Japanese vocabulary (plus ゐ wi and ゑ we, which are obsolete, and ゔ vu, which doesn't occur in native or Sinitic vocabulary but has its own kana). [ɕe ʨe ʥe] don't occur in native vocabulary because they would phonemically be /sje tje dje/, and /je/ doesn't exist in native vocabulary (it merged with /e/ long ago, which is why the word for yen (円) is "en"). [ɸa] doesn't occur in native vocabulary because [ɸ] is an allophone of /h/ before /u/.
Post Reply