Talk:Japanese phonology

Spinning off an onbin article

Since the article is getting on the longer side now, I was thinking it might make sense to create a separate article about onbin that could get more into the details of it as a historical process (rather than just part of the current language). I started a draft here: User:Urszag/Onbin but I wanted to share this suggestion now since I see @Mazamadao has been working now on expanding the onbin section. My thought would be that specific examples of lexical onbin, such as those currently listed in the "-hito" section (shirōto, etc.) are more a matter of diachronic change than synchronic phonology, and so could be moved into a new onbin article (I would also move dialect verb forms, since this article is not intended to comprehensively describe dialects other than standard Tokyo-based Japanese), while a summary of onbin changes, and examples of grammatical onbin in standard Japanese inflection, should remain in this article. Does anyone object to moving that material once a new article is created? Urszag (talk) 08:28, 8 September 2024 (UTC)

I published Onbin as its own article. Since nobody commented, I went ahead and moved the specific examples of onbin in compounds of hito/-bito to that page.--Urszag (talk) 06:41, 13 September 2024 (UTC)

Where are the consonants for じ・じゃ・じゅ・じょ and し・しゃ・しゅ・しょ?

They are part of Japanese Language but can't be found in the consonants table anywhere. Why? 2804:14C:120:2EC6:1CF3:C0BF:3BF6:1BA3 (talk) 05:19, 25 December 2024 (UTC)

They're there. They're in parentheses because they are not phonemic (in the analysis adopted by the article). Nardog (talk) 06:08, 25 December 2024 (UTC)

I thought it was really confusing for a layperson to understand where they were. The article states for instance that the Voiceless alveolo-palatal fricative is present in the language but doesn't make clear which sound in Japanese it corresponds to with examples.

I'd like to suggest the adoption of a table like the one found on the consonants section of this article Standard Chinese phonology. In this table examples are given showing which consonants in Mandarin corresponds to each sound in the IPA, withe example words and even pronunciation and comparison to English. That would be sure to make it more understandable for people not so familiar with the IPA transcription. I, for instance, couldn't tell which sound the Voiceless alveolo-palatal fricative was, because I expected the sound instead to be Voiceless postalveolar fricative. 2804:14C:120:2EC6:1CF3:C0BF:3BF6:1BA3 (talk) 19:29, 25 December 2024 (UTC)

[ɸ] to [ç] phonetic shift clarification

Under this subheading, near the bottom of the second-to-last paragraph, it is explained that "The labial fricative could be found before all vowels up through Late Middle Japanese, but was eventually debuccalized to before any vowel other than , resulting in the modern Japanese phoneme." I believe the explanation would benefit from more clarity on how shifted to before , especially as it currently seems to incorrectly suggest that is realised as rather than before .

Unfortunately I do not know, so I am requesting that somebody more well-versed than me in this topic could please provide the clarification. I realise there is a possibility this may seem lazy if the standard procedure would be to research this myself and either make the edit or suggest the exact edit with sources, so I apologise if this is the case. However I thought in any case it would be more helpful to bring it up than do nothing, so I decided to open this topic. Jai (talk) 18:58, 30 January 2025 (UTC)

I've edited the description of the sound change to account for [ç]. I don't recall if any sources specifically talk about the steps in its evolution.--Urszag (talk) 23:45, 30 January 2025 (UTC)

Labrune 2012 as a source for details on devoicing of non-high vowels

@Mazamadao: Thanks for your edits to the article! I haven't yet read the sources that Labrune (2012) cites about devoicing of non-high vowels, but I'm uneasy with all of the restrictions that Labrune postulates and with the statistic that /e/ is least likely to be devoiced. Fujimoto (2015), a more recent and perhaps more comprehensive overview of this topic, says on page 169 that a study by Maekawa and Kikuchi in 2005 found devoicing rates of 2.10%, 3.31%, and 3.45% respectively for /a, e, o/, which would make /a/ and not /e/ the least likely to undergo devoicing (Fujimoto also mentions the conflicting studies that found /e/ to be least likely, and concludes that "it is not clear" which is most likely to devoice). We could go over this contradictory information in the article, but the article is already getting a bit lengthy and I'm not sure it's useful to be more detailed here rather than just avoiding commenting on this specific question, which doesn't seem of clear importance. I'm also not convinced the restrictions mentioned by Labrune are clearly established enough to warrant our repeating them. The restriction "followed by the same vowel in the following mora" is not claimed to apply to /e/ by Labrune: in fact, Labrune's examples for e̥ are counterexamples to that (sekkaku and keshō). Counterexamples for [ḁ] and [o̥] are mentioned by Fujimoto, namely [kḁkɯte:] 'local train' and [hanayakḁ] 'brilliant' and [te:ʃo̥kɯ] 'steady job'. So even if devoicing is more common when the same vowel follows, this doesn't seem to be an inviolable constraint, and I'm not sure how helpful Labrune's "in theory" wording is. Urszag (talk) 09:27, 7 February 2025 (UTC)

Feel free to cite other sources that say otherwise. I won't editorialize and will simply paraphrase what I've read. I'm not the one to direct these questions too, I'm just a book reader. Mazamadao (talk) 11:33, 7 February 2025 (UTC)

I added the references mentioned above. Does "Feel free to cite other sources that say otherwise" mean that you're opposed to removing the sentence about the relative frequency of devoicing for /a e o/ from the article?--Urszag (talk) 22:29, 9 February 2025 (UTC)

Kochetov 2014 and Kochetov 2018 as sources on articulation of coronal consonants

There's a disagreement between me and Mazamadao about whether to include these sources. You can see the arguments for and against inclusion in the article history and on my Talk page. Responding to Mazamadao's last comment there: the term "laminal denti-alveolar" has never been my own unique understanding, but is Kochetov's own description of the place of articulation of consonants like /t/. Urszag (talk) 10:13, 8 February 2025 (UTC)

There's no such thing as "Kochetov's own description." Kochetov doesn't offer any unique description, but merely cites Vance and Okada. If the first-hand accounts are already cited in this article, why would you need someone's second-hand account? To be clear, whenever I read "dental" as regarding /t, d/, I tend to think of it as somewhere that can be characterized as "denti-alveolar", but that does not give me license to change the wording however I want. I'd like you to demonstrate clearly that Kochetov has indepedently verified "denti-alveolar", not just citing someone else who doesn't use that term.Mazamadao (talk) 10:31, 8 February 2025 (UTC)

And to be extra clear, it's not just "denti-alveolar", but also "laminal".Mazamadao (talk) 10:33, 8 February 2025 (UTC)

Here are the relevant passages in Kochetov's Voicing and Tongue-Palate Contact Differences in Japanese Obstruents (2014). I added bold type.

"The results of the study are largely consistent with previously reported voicing differences (except for Dagenais et al 's 1994 results for stops). Specifically, we found that Japanese voiced alveolar /d/ was consistently produced with less linguopalatal contact than voiceless /t/. The degree of contact difference was primarily due to the relative anterior-to-posterior width of the closure at the alveolar ridge (narrower for /d/), but was also affected by the lateral contact in the palatal region (less for /d/). Unlike for /t/, the anterior edge of the closure for /d/ was slightly less advanced, and the closure was often less occluded (especially, but not exclusively, before high vowels). Similar results have been reported for alveolar stops in British English, German, Hindi, and Norwegian (Dixit 1990, Moon and Simonsen 1997, Fuchs and Perrier 2003, Fuchs 2005). Previous EPG studies of Japanese showed a similar difference in the degree of contact (Shibata 1968, Fujii 1970), likely reflecting differences in articulatory force (Matsumura et al 1994, Wakumoto et al 1998). Together with these studies, our results thus show that Japanese voiced alveolar /d/ is also subject to lenition (contra to previous descriptive phonetic accounts: Vance 1987, Okada 1999), yet the magnitude of this process is likely to be smaller than the lenition of velar /g/ and labial /b/, and likely less auditorily salient.

The observed contact differences between Japanese /t/ and /d/ are relatively robust, which is possibly due to the overall high degree of contact for these consonants that are commonly described as laminal denti-alveolars (Vance 1987, p. 18, Okada 1999). Similarly clear differences were observed in Hindi and Norwegian /t, d/, which are also laminal denti-alveolars (Dixit 1990, Moon and Simonsen 1997)."

Basically, Kochetov doesn't independently claim that /t, d/ are "laminal denti-alveolars", but only points out that Vance and Okada say so. Vance (1987) and Okada (1999) are cited in this article. If you want to cite them, you are free to do so, but everybody else is also free to verify whether the claim is found in the sources. Okada uses the term "dental" and I don't see any reason why I have to rely on Kochetov's interpretation that it means "denti-alveolar", because unless Kochetov can independently come up with their own description, however they personally interpret Okada's "dental" is irrelevant. What Okada says is what they say, it shouldn't be filtered through somebody else's mouth.

As for Shibata (1968), Fujii (1970), Matsumura et al (1994), Wakumoto et al (1998), these are about how much /d/ makes less palatal contact than /t/, and how /d/ is subject to lenition, which are irrelevant to the claim in question, that /t/, /d/ AND /n/ are "laminal denti-alveolar."Mazamadao (talk) 11:12, 8 February 2025 (UTC)

Template:Ping "There are a number of reasons to cite Kochetov in addition to citing Vance and Okada. For one thing, Kochetov's articles are accessible via jstage whereas Vance and Okada's books are not; Kochetov's articles are also more recent sources. Furthermore, while Kochetov cites Vance and Okada in that particular part of the article, my point is that it's clear Kochetov overall is familiar with more sources of information on this topic than just Vance and Okada's descriptions, so there's no reason to assume that Kochetov's description is solely an "interpretation" (you seem to be suggesting a misinterpretation?) of these two sources." These are troublesome reasoning. Apparently if it's free and more recent, it should be included? To what end exactly? And again, why would you need a second-hand account if a first-one already exists, accessible, albeit not completely free and not recent? You're grasping at straws with that whole "Kochetov cites Vance and Okada in that particular part of the article" despite that's where the only instance of "laminal denti-alveolar" is found. I have to caution you again, it's not just the "denti-alveolar", which I did find on page 65, it's also the "laminal" part. You can only cite Kochetov to the extent that your claim is only "denti-alveolar." The word "laminal" is literally not anywhere else in that entire paper.Mazamadao (talk) 11:49, 8 February 2025 (UTC)

Thank you for the additional comments. So you are OK with citing Kochetov if that portion of the sentence is changed from "laminal denti-alveolar" to "denti-alveolar"?--Urszag (talk) 20:26, 8 February 2025 (UTC)

Yes. When I checked Kochetov's paper, the first thing I did was literally looking up "laminal". I've never found the distinction between "dental" and "denti-alveolar" that interesting to begin with. Mazamadao (talk) 05:03, 9 February 2025 (UTC)

Gemination

There is a lot, about gemination, in this article, that doesn't appear to make sense.

"and can be transcribed phonetically as two occurrences of the same consonant phone in sequence: a geminate plosive or affricate is pronounced with just one release, so the first portion of such a geminate may be transcribed as an unreleased stop.[93]"

...or, in other words, a glottal stop.

"Geminate consonants can also be phonetically transcribed with a length mark, as in [ipːai], but this notation obscures mora boundaries."

...as well as being completely inaccurate, as that isn't how it is pronounced. It is, indeed, physically impossible to pronounce [pː], or lengthen the pronunciation of any other plosive/stop.

"Vance transcribes Japanese geminates with two length markers, e.g. [sɑ̃mːːɑi], [ipːːɑi], and refers to them as "extra-long" consonants, on the grounds that there is no acoustic boundary between two halves of a geminate.[94]"

This is clear and obvious nonsense. さんまい (sanmai) can be pronounced without an acoustic boundary between two halves of a geminate ...though it most certainly is sometimes pronounced with a clear acoustic boundary. (also, it doesn't involve 'っ' or /Q/, and is therefore irrelevant) For いっぱい (ippai), however, this is neither true, nor physically possible. With non-stops/plosives, such as with いっさい (issai), it would be perfectly physically possible ...but isn't true of how it is actually pronounced. Also, even if none of that were true, marking it with two lengthening marks, suggests it is three times the length, rather than just double ...which conflicts with reality.

"Alternatively, it has been suggested that the underlying phonemic representation of /Q/ might be a glottal stop /ʔ/—despite the fact that phonetically, it is not always a stop, and is usually not glottal"

What evidence is there, for 'っ' ever not being a glottal stop?

"For example, [ʔ] can be found at the end of an exclamation"

Was this meant to say /Q/ or っ, rather than [ʔ]? Either is true. You do sometimes find a 'っ', at the end of an exclamation ...and there are several examples, of when a glottal stop can be found at the end of a word or sentence.

"or before a sonorant in forms with emphatic gemination"

There is no instance of where 'っ' is used to geminate a sonorant, in Japanese, where it isn't a glottal stop. Sonorants can be geminated by simply lengthening the consonant sound, but this is never done in Japanese. All vowels are geminated using a glottal stop, aside from when ん is followed by a 'n' or 'm' syllable. (e.g. min'na/みんな or うんめい/unmei) In other words, every instance of gemination using 'っ' (or /Q/, as some like to call it), is using a glottal stop.

"or the form it takes when it is not possible for it to share its place and manner of articulation with a following obstruent.[99]"

There is no instance where it takes any form, other than a glottal stop, that I have ever seen, heard, or heard about.
Ever. 185.113.97.24 (talk) 21:24, 3 May 2025 (UTC)

There is a plain articulatory difference between an unreleased stop and a glottal stop: the latter involves the glottis, whereas the first doesn't (assuming it is a consonant like [p̚] with a non-glottal POA). In English, syllable-final stop consonants tend to be both unreleased and glottalized, but these are not the same thing. There are languages such as Arabic where sequences such as /ʔt/ and /tt/ are both possible and contrastive. If long plosives were impossible, the 1999 Handbook of the IPA wouldn't give [tː] as an example of a long consonant (page 22).

I don't particularly like Vance's "ːː" notation or think it is necessary, but the idea behind it is that the moraic consonants N and Q are longer than syllable-initial consonants, meaning moraic/coda [m] in Japanese can be transcribed as [mː], in contrast to non-moraic onset [m]. Since a geminate phonologically contains both a moraic consonant and a following syllable-initial consonant, geminate /Nm/ can be imagined to be [mː] + [m] = [mːm] = [mːː].

Regarding the pronunciation of っ before sonorant consonants: Kitaoka 2017:5 and Morimoto 2020:46–51 both say that the production of a glottal stop in this context is optional, not mandatory. The latter provides spectrograms showing pronunciations with no interruption of voicing, and also mentions that laryngealization is not acoustically detectable in many cases for geminate /r/, and is never detectable for geminate /s/.--Urszag (talk) 00:23, 4 May 2025 (UTC)

"There is a plain articulatory difference between an unreleased stop and a glottal stop/.../"

Can you provide some explanation/clarification of this? I'm not at all sure, that I understand what you're saying ...and can you provide some audio evidence?

"but the idea behind it is that the moraic consonants N and Q are longer than syllable-initial consonants"

I see no evidence of that, in any way. I've never heard a /N/ or /Q/, that is noticeably longer, than syllable-initial consonants. Certainly not twice as long.

As for your response to っ before sonorant consonants... Can you clarify what you said? What is claimed to happen, in instances where a glottal stop isn't present? And can they/you show any evidence, of っ before a sonorant consonant, that is pronounced without a glottal stop? 185.113.97.24 (talk) 01:38, 4 May 2025 (UTC)

Have you looked at the relevant parts of the sources? They're both linked from the article. If you can identify a specific way Morimoto's evidence is lacking, that would be helpful.--Urszag (talk) 17:27, 4 May 2025 (UTC)

That is not an explanation or clarification, and I can hardly expect an article in a scientific journal, or a PhD thesis (why is a mere thesis, counted as a RS?), to use clearer and simpler language and explanations. Also, how do they determine what is, or isn't, a glottal stop?

"If you can identify a specific way Morimoto's evidence is lacking"

Tell/show me what it is, and I'll be able to give a response to it. 185.113.97.24 (talk) 17:53, 4 May 2025 (UTC)

I'm not enough of an expert to give you a short, simple, clear explanation. Here's an attempt to clarify this, though.

An unreleased stop is made by holding articulators in a particular position without releasing it. For a bilabial stop, the articulators are the upper and lower lips. So unreleased [p̚] is made by closing the lips, thereby obstructing airflow at the lips. Although it is called "unreleased", there is eventually a release: the release comes at the end of the second consonant in the geminate sequence [p̚p]. In practice, the difference between [p] and [p̚p] is that the release is later in the second, i.e. the lips are held closed for a longer amount of time in in [ap̚pa] (which can also be transcribed [apːa]) than they are in [apa].

Holding your lips closed for a longer period of time has no inherent connection to a glottal stop. The sound [ʔ] is articulated in the larynx by moving the vocal folds very close together. This obstructs airflow in the larynx, but often does not completely stop vibration of the vocal folds, which means [ʔ] is often at least partially voiced. (Hence the name "glottal stop" is a bit of a misnomer, as it often does not involve the full obstruction of airflow that gives "stop" consonants their name.) See Voicing of glottal consonants and non-modal vowels for more information on the phonetics of glottal stops.

The vocal folds are normally moderately constricted and vibrating during the pronunciation of voiced sounds (including vowels and voiced consonants), and very constricted (to the point that their vibration may be interrupted) during the pronunciation of a glottal stop. In contrast, during the pronunciation of most voiceless consonants, the vocal folds are often opened or spread.

The degree of closure of the vocal folds can be indirectly measured by a technique called electroglottography, where electrodes are placed on the neck. The level of impedance serves as a proxy measure for the degree of vocal fold contact area: a decrease in impedance (or resistance) corresponds to an increase in vocal fold contact (closure/constriction). This is explained in this article: Electroglottography in Medical Diagnostics of Vocal Tract Pathologies: A Systematic Review. I found this article "Electrical Glottography" which has a chart on the last page showing how a glottal stop shows up on an electrical glottogram: it appears as an area of lower resistance than the surrounding vowel sounds, in contrast to the voiceless stop [k], which appears as an area of higher resistance.

You can see pictures and glottogram charts in Fujimoto, Maekawa & Funatsu (2010) that show that Japanese geminate consonants can be pronounced with no appreciable closure of the vocal folds. In contrast, they are sometimes pronounced with the vocal folds opened or spread, especially during the pronunciation of voiceless fricative sounds such as [ss] and [ɕɕ].

From what I can see, Morimoto 2020 relied on acoustic measures to try to detect laryngeal constriction. This can give a clear positive result in cases where voicing was obstructed. I'm not sure how reliable it is at avoiding false negatives in cases where voicing is uninterrupted. I agree that a thesis by itself is not a strong source, especially when reporting primary research. However, using theses and primary sources as references is not forbidden in the absence of anything better. Per WP:SCHOLARSHIP, one factor that may count in favor of using a thesis is citations from later literature: Google Scholar shows that Morimoto 2020 has been cited by some journal-published articles, such as "Japanese speakers can infer specific sub-lexicons using phonotactic cues" (S Kawahara, G Kumagai - Linguistics Vanguard, 2024). Akamatsu 1997 choses to mostly avoid this topic, but briefly mentions [ll] as being possible in the pronunciation of the expression arre, maa "Oh dear!" (page 154, also footnote 439): this transcription supports the statement that geminate -rr- in Japanese does not necessarily involve a glottal stop.--Urszag (talk) 00:04, 18 May 2025 (UTC)

You should have made that a bit simpler, with more explanations of technical terms, but... That did clarify a lot! Thank you!
I accept that you're correct, and I was mistaken.
So essentially, 'っ' is an... "abruption of pronounciation" ...which can be in the form of a glottal stop, but can also be done in other ways.
I had come to understand, that "glottal stop" meant "abruption of pronounciation", but that's apparently wrong.
Your explanation, makes it make a lot more sense, to be honest.

"but briefly mentions [ll] as being possible in the pronunciation of the expression arre, maa "Oh dear!""

Arre, maa? I've never heard of such an expression. You often come across "Ara, maa", "ara ara", "are" etc, certainly, but never "arra" (whether in isolation or in combination with other word[s]), in any way, shape, or form. Araa (or, more accurately, ara~), at times, but never arra
...and looking it up, the only thing I can find, that comes close, is アッラー as one of the possible transcriptions of the Arabic word Allah. 185.113.99.207 (talk) 11:33, 18 May 2025 (UTC)

Right, a glottal stop is specifically an obstruction made by closing the glottis (the space between the vocal folds), but there are other ways to obstruct airflow: [p̚] uses the lips, [t̚] uses the tongue and teeth (or alveolar ridge behind the teeth), [k̚] uses the tongue and the soft palate.

"Arre, maa" does not seem to be a well-established spelling, just a way to refer to a particular variant pronunciation of the exclamation あれまあ. I see some results for "あっれまあ！", "あっれ，まあ！" etc. on Google. What Akamatsu says on page 154 is: “[ll], but not [ɾɾ], may marginally occur in an exclamatory expression arre, maa (rr here being a hypothetical spelling) meaning ‘Oh, dear!’,⁴³⁹ but may safely be left out of account for the purpose of this book. Even this exclamatory expression can always be alternatively pronounced [aɾe maa] accompanied by appropriate facial expression.” Akamatsu’s footnote 439, on page 333, reads: “This example is cited by Kawakami (1977, p. 51). He says that [ll], as in this example, will occur in the speech of even those Japanese speakers who normally do not pronounce [l] before vowels. That [ll] would otherwise be rare in Japanese can be seen, it seems to me, from the fact that, for example, Corelli (an Italian composer) is normally pronounced, as a loanword, [koɾeɾɯɾ’i] rather than [koɾel’l’i].” I’m not sure whether Akamatsu’s comment about loaned names like Corelli normally being pronounced with [ɾɯɾ] is accurate for most present-day Japanese speakers.--Urszag (talk) 18:24, 18 May 2025 (UTC)

Hm... further attempts to find "arre", does show results. It appears that it is sometimes said/written, though I doubt it's all that common. I'm a bit doubtful, that "rr" wouldn't involve an "abruption of pronounciation", but as I've never heard someone say "arre"... 185.113.99.207 (talk) 19:24, 18 May 2025 (UTC)

I suggest you consult any textbook on the relevant field before making such sweeping claims. It would tell you what makes a plosive (or any occlusive) a plosive/occlusive is the occlusion, and may lack an approach or release phase. Nardog (talk) 00:16, 17 May 2025 (UTC)

So what you're saying, essentially, is that you cannot verify the claims you are making. Also, please note WP:RSPRIMARY. (and MOS:JARGON) 185.113.97.151 (talk) 05:27, 17 May 2025 (UTC)