Help talk:IPA/Japanese/Archive 3

Archive 1

Archive 2

Archive 3

Word-final devoicing

@Libhye, IvanScrooge98, and Aeusoes1: Do we even need to mark word-final deovicing? First, according to typical summary descriptions like Labrune (2012: 34), word-final devoicing of high vowels occurs (or is obligatory?) only when utterance-final. I'm not totally sure about positions before a voiceless consonant, but when followed by a voiced sound, as in common particles like o/ga/ni/wa, a word-final high vowel subject to devoicing is guaranteed to not undergo devoicing. Second, according to Hasegawa (1999) which Libhye referred to, word-final devoicing is only optional and is found in both accentless and accented words (see § 3.2). Third, desu/masu, which do seem to undergo devoicing consistently in utterance-final positions and are most often utterance-final, are function words (copula/suffix) so very unlikely to be used in IPA-ja transcriptions (even if they were used we could single them out). (EDIT: Elision and Bach no Senritsu o Yoru ni Kiita Sei Desu are the only articles currently using IPA-ja with desu or masu in it.) Finally, non-high vowels also undergo devoicing in a less consistent manner, but we are not marking devoicing of them. So it appears to me we only need mark devoicing in word-medial high vowels surrounded by voiceless consonants, except perhaps phrase-final desu/masu. Nardog (talk) 03:05, 24 September 2019 (UTC)

You make a compelling case for not marking word-final devoicing. If we don't mark it, we probably shouldn't mention it here. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 04:17, 24 September 2019 (UTC)

Then I think we should reword the note. I wouldn’t remove it straight away. 〜イヴァンスクルージ九十八［IvanScrooge98］（会話） 06:23, 24 September 2019 (UTC)

I think this is a bad idea for the reason that word-final devoicing of high short vowels is obligatory in some words and prohibited in others, and the distribution is not predictable. In other words, the distinction between voiced and voiceless word-final high short vowels is phonemic, and not transcribing all phonemic distinctions should be a huge no-no.

Note that Hasegawa does not say that word-final devoicing is only optional, but that it is obligatory in some words. She says that désu and -másu are examples of words/suffixes where it is obligatory, not that they make up the entire list. She also indicates that these two always undergo devoicing; she doesn't say a word about phrase-final or utterance-final positions. (Labrune might say something about this, but if we are to learn anything from Hasegawa's article, surely it is that people who make generalisations about Japanese vowel devoicing tend to be completely wrong – she mentions no less than three examples of this.) Also note that even in those cases where word-final devoicing is optional, the two pronunciations are not necessarily equal: Hasegawa says that the word for ‘autum’ is normally pronounced [aꜜkʲi̥].

The voiced : voiceless distinction in non-high vowels isn't phonemic, so that is a different issue. Libhye (talk) 03:24, 26 September 2019 (UTC)

@Libhye: If it is phonemic, then can you name a minimal pair that demonstrates the phonemicity? And if it is unpredictable, then how would you suggest to source each pronunciation? It is not reflected in kana spelling, and general dictionaries like Daijirin and Daijisen only indicate the position or absence of accent, not devoicing. And not every word transcribed using IPA-ja is included in accent dictionaries.

She says that désu and -másu are examples of words/suffixes where it is obligatory, not that they make up the entire list. Then what are the rest of them? Are there any content words where word-final devoicing is obligatory?

I'm pretty sure the final vowel in aki etc. doesn't get devoiced when followed by a voiced sound. The waveform on p. 36 of Labrune illustrates this. Even desu/masu aren't always devoiced. Vance (2008: 212–3) says, "If the last syllable in a sentence contains a short high vowel preceded by a voiceless consonant but has to carry the intonation for a question, the vowel doesn't devoice. As an example, consider the question /ikimaꜜsu⤴/ 行きます？ '(Will you) go?'." He also says devoicing of /u/ in desu or masu followed by a pause (i.e. utterance-final) is just "quite consistent" and "usual" for "most Tokyo speakers", not that it's mandatory.

Shaw & Kawahara (2018) mention Fujimoto (2015) as "a recent, comprehensive overview" of studies on vowel devoicing in Japanese. So I took a look at it, and as it turns out, it gives a picture of vowel devoicing in Japanese that is much more variable than I envisaged. I highlight:

Contrary to the description that devoicing is obligatory in Tokyo Japanese, its actual occurrence diminishes due to many factors such as consonantal environment, accent, speech rate, and dialects. (167)

From these results, it is reasonable to categorize the consonantal conditions of high vowel devoicing into two types: typical and atypical. In the 'typical' consonantal conditions devoicing occurs systematically and regularly in Tokyo speakers as described by many traditional studies. They are 'St-St', 'St-Fr (except for /h/)' and 'Fr-St'. In the 'atypical' consonantal conditions, devoicing occurs randomly with greater inter-speaker variation. They are 'Fr-Fr' and 'St/Af/Fr-/h/'. Within atypical consonantal conditions, probability of devoicing is lowest when the target vowel is followed by an /h/. (172)

As was shown in many studies, devoicing between a voiceless consonant and a pause does not consistently occur. Interpersonal variation may be one cause. More importantly, however, the above results suggest that devoicing is more likely to occur phrase-finally than word-finally. In phrase-final position, function words such as desu and masu often appear, in which devoicing is very common. In word-final position, in contrast, word frequency effect is not much expected. In addition, speakers may pronounce words elaborately when uttered in isolation, which may decrease the devoicing rate word-finally. (185)

In sum, word/phrase-final devoicing is not as systematic as was traditionally described. Devoicing is plausibly more frequent phrase-finally than word-finally. The effects of word frequency, word type (i.e. content vs. function words), accentual patterns and intonation should all be taken into account. Further investigation is essential in order to clarify the details of word/phrase-final devoicing. (186)

So my view that word-final devoicing need not be marked in our transcriptions except for phrase-final desu and masu does not waver, and is reinforced. Fujimoto is substantiated by far more previously-published empirical studies than Hasegawa, which cites hardly any. Nardog (talk) 18:18, 26 September 2019 (UTC)

Minimal pairs are not needed to establish phonemicity, only unpredictability. The near-minimal pair -másu : kásu, where, according to Hasegawa, devoicing is obligatory in the first but optional in the second in declarative intonation (I should have mentioned that caveat), is sufficient to establish phonemicity.

Then what are the rest of them? She doesn't say, but implies that other examples exist by using ‘e.g.’.

As for sources, there aren't any. The right way to deal with such a situation is to leave transcribing to native speakers of Standard Japanese. The last thing we should do is invent a system where devoiced high vowels are sometimes indicated and sometimes not, based not on any attribute of Japanese standard pronunciation but merely on the fact that we are ignorant of the pronunciation of the words in question. If we don't know how a word is pronounced, we simply shouldn't transcribe it. Having read Hasegawa, I believe the risk that we are dealing with a phonemic distinction is too high for us to take the chance there is no phonemic distinction and risk giving transcriptions that are phonemically incorrect.

As Fujimoto says, [f]urther investigation is essential in order to clarify the details of word/phrase-final devoicing. For now, we simply don't know. Is the distinction phonemic? It is a distinct possibility (it is implied by Hasegawa's data, so those would have to be incorrect for the distinction not to be phonemic). Is it non-phonemic? Also a possibility. We don't know for certain either way. But what we do know for certain is that if we treat it as non-phonemic, we run the risk of giving phonemically incorrect transcriptions, while if we treat it as phonemic, we run no such risk. We don't get to take that substantial risk on behalf of our readers. Libhye (talk) 21:50, 26 September 2019 (UTC)

This sounds like slipshod phonological analysis. Phonemic or not, we should not be implementing a transcripton scheme that depends so thoroughly on native speakers, especially for something that sounds a lot like it's going to vary from speaker to speaker. We don't want a system where devoiced high vowels are sometimes indicated and sometimes not, based not on any attribute of Japanese pronunciation but merely because we haven't found native speakers to confirm or correct given transcriptions. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 22:33, 26 September 2019 (UTC)

The question is, why would we want a such a system (where devoiced high vowels are sometimes indicated and sometimes not) at all – no matter what it's based on? It's not like devoicing is predictable between two voiceless consonants either. Since devoicing is unpredictable both between two voiceless consonants and word-finally after a voiceless consonant, we should either transcribe it in both positions or refrain from transcribing it in both positions. Transcribing it in some cases and not in others is unacceptably misleading. Libhye (talk) 05:42, 28 September 2019 (UTC)

Minimal pairs are of course required to establish phonemicity—unpredictability can just mean allophonic variation—unless you've got a large corpus where the feature appears consistently in certain lexemes and not in others. And Fujimoto shows that it isn't consistent, citing corpus studies and experiments. If your interpretation of Hasegawa—which is a proceeding paper from 1999, unlike Fujimoto which is an overview of previous research included in a 2015 anthology handbook—is enough to establish phonemicity, then why isn't it established? If this was about a little-studied language, things might be a little different, but Japanese is a well-documented language whose study goes back centuries (Lyman's law was discovered in the 1700s). So why don't newer, comprehensive descriptions of Japanese phonology like Labrune and Vance mention its being phonemic? Not to mention Hasegawa doesn't even discuss the phonemic status of the feature; the word phonemic, distinctive or contrastive or its derived form is nowhere to be found in her discussion of word-final (or any) vowel devoicing. To infer such a conclusion is patent OR.

Relying on native speakers' self-reports runs directly counter to one of Wikipedia's fundamental policies, that it requires content to be verifiable with reliable sources. Anyone can claim to be a native speaker of "Standard Japanese" (not to mention there aren't many such people), and who is there to verify them? We don't record something just because it might be phonemic. That's WP:CRYSTAL and WP:OR. Nardog (talk) 00:38, 27 September 2019 (UTC)

Why shouldn't we record a phonetic distinction for the reason that we're not sure it's non-phonemic? Seems like a good reason to record it. This is a talk page, so the ban on OR doesn't apply here, and we don't provide phonemic transcriptions but phonetic ones, so including non-phonemic dictinctions is not in violation of WP:OR. If we use voiceless word-final high vowels in some phonetic transcriptions and voiced ones in others, we're not in any way, shape or form saying that the distinction is phonemic. The question is rather, how could we possibly justify transcribing voiceless high vowels in some cases and not in others in phonetic transcription? We should either provide all voiceless high vowels or none of them.

Hasegawa does claim that ‘the feature appears consistently in certain lexemes’. Anyway, you're missing my main point, which is that the distinction might be phonemic and that that's reason enough to transcribe it (a view with which you disagree). This part of Japanese phonology simply hasn't been exhaustively described. Nowhere have I implied that Hasegawa says anything about phonemicity; I'm talking exclusively about her data (and of course that is OR, but I'm not suggesting putting in an article that the distinction is phonemic, only providing arguments for how we should go about giving phonetic transcriptions). As for no one having said in so many words that the distinction is phonemic, that's hardly surprising giving the evidence, which continues to confuse everyone. The point is that there's a substantial chance it might be phonemic and that that chance should influence how we decide to transcribe Standard Japanese phonetically.

You seem to think the onus is on me to provide a reason why devoiced high vowels should be transcribed also in word-final position. But surely the normal thing to do is either transcribe these vowels always, or never transcribe them, and the onus is on you to provide a sensible reason why such vowels should be transcribed when they occur between two voiceless consonants but not when they occur word-finally. Transcribing them between voiceless consonants runs into the exact same problems as transcribing them word-finally: a short high vowel may be voiceless or voiced between voiceless consonants, multiple pronunciations sometimes being correct for one word, and there are no sources whatsoever. Singling out the word-final position makes no sense. It would be better to never transcribe devoicing than to leave it out only sometimes on flimsy grounds.

Requiring phonetic transcriptions to be sourced would lead to the removal of nearly all such transcriptions from Wikipedia, to the great detriment of the encyclopaedia. There is currently a working consensus that they don't need to be sourced, but this consensus would unravel as soon as an appreciable number of uninvolved editors had their say. It is therefore of utmost importance that no editor involved in phonetic transcription ever suggests that such transcriptions need sources. Libhye (talk) 05:42, 28 September 2019 (UTC)

While we don't go to the level of requiring each individual word have a source backing up a phonetic transcription, especially for languages with relatively predictable pronunciations, it is simply not the case that we are lax on our OR standards for pronunciations because it would be hard otherwise. That is most certainly not how things work here.

The impression that I got was that non-final devoicing was, comparatively speaking, more predictable. Is this incorrect? — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 02:23, 29 September 2019 (UTC)

@Aeusoes1: According to Fujimoto, no, it is not incorrect—at least in certain environments. The table on p. 179 sums this up. I think we should just mark devoicing only in the environments "typical" in her designation and deal with the others in a footnote in the guide. Nardog (talk) 17:01, 29 September 2019 (UTC)

There can be no doubt that the average Wikipedian would consider our current practice of not requiring a source for every pronunciation to be in breach of WP:OR. Saying that ‘it is simply not the case that we are lax on our OR standards’ and ‘[t]hat is certainly not how things work here’ wouldn't convince them. Every time someone brings the existence of a bad pronunciation editor to the attention of the wider community, there are voices saying that every pronunciation needs to be sourced, and only editors already involved in providing and editing pronunciations contradict them. At the same time, the vast majority of phonetic transcriptions on Wikipedia are not accompanied by a reference. Therefore, bringing up the need for sourcing transcriptions, particularly in places where uninvolved editors may see it, carries with it the risk of utterly destroying English Wikipedia's coverage of pronunciation. It has already been close more than once, and if editors don't watch their mouths, it is only a question of time before it is going to happen.

Anyway, back to the question at hand: Even if it is true that non-final devoicing is, ‘comparative speaking, more predictable’, that is not a valid argument for transcribing devoicing of high vowels only sometimes in phonetic transcription (does anyone else do this?). If anything, the less predictable a phonetic feature is, the more important it is to transcribe it, so a valid argument would be the exact opposite: since word-final devoicing is less predictable, it should be transcribed. We already transcribe palatalisation of consonants before /a, u, o/ but not before /i/, precisely because palatalisation is unpredictable in the former cases but predictable in the latter case. Why should devoicing be treated in the opposite fashion?

I remain, however, unconvinced that the predictability of non-final devoicing is sufficiently high for us to treat devoicing in the two positions differently. Some sources say that accented high vowels are never devoiced, others that they are frequently devoiced and that devoicing is obligatory in some cases. Some sources say that it is rare to have two consecutive morae with a devoiced vowel, others that it is commonplace. Some sources say that it is impossible to have three consecutive morae with a devoiced vowel, others that there is nothing special about it. In short, if we take all sources into consideration, we are as clueless as to where to devoice in non-final position as in final position. If we artificially confine ourselves to one source, things might look different, but that is no way to deal with sources. Libhye (talk) 11:16, 30 September 2019 (UTC)

Per WP:CHALLENGE: Attribute all quotations and any material whose verifiability is challenged or likely to be challenged to a reliable, published source using an inline citation. Pronunciations that are backed up indirectly by sources that discuss sound-to-grapheme correspondences or language-specific phonology, are extremely unlikely to be challenged. Even so, they are still subject to this same scrutiny; if someone challenges a given pronunciation on WP:OR grounds, a source for that specific pronunciation satisfies WP:OR concerns. This "working consensus" that you're talking about is not at risk of being changed by phantom mobs of Wikipedians demanding explicit in-text citations for individual pronunciations. That's just absurd.

You and I seem to be using "predictable" in different ways. When I say predictable, I mean that a given word will feature word-final devoicing consistently from speaker to speaker in a given dialect, preferably a standard variety. You seem to mean it in the sense that it is not phonemic. Your sense of predictability is irrelevant to our transcriptions because it's not a phonemic transcription. Predictability in the sense that I have been talking about it is vital because a feature that varies in realization or distribution from speaker to speaker isn't something that we can reliably transcribe.

I trust Nardog's assessment and agree that we should mark devoicing only in the environments that Fujimoto designates "typical". — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 15:28, 30 September 2019 (UTC)

The question isn't what actually satisfies WP:OR concerns, but what people think satisfies, and doesn't satisfy, such concerns. Wikipedia is close to being a democracy, after all. The editors I'm talking about aren't phantom editors, but real editors that have participated in real discussions. They make up most of the uninvolved editors who participate in such discussions, and they do want to remove all pronunciations that aren't explicitly cited. If the number of uninvolved editors who participate in such discussions rises by only a few, and they hold the same opinions as the other uninvolved editors, that will necessarily lead to a consensus that every pronunciation needs to be individually sourced. To call this prospect absurd makes no sense: obviously, the number of uninvolved editors participating could easily reach the relevant threshold the next time someone brings up an editor making bad pronunciation edits at WP:ANI. These editors, having established the new consensus, might challenge individual pronunciations before they remove them – or maybe they will just remove them. Either way, the pronunciation-editing community will not be able to come up with sources for the vast majority of unreferenced pronunciations, and so most pronunciations on English Wikipedia will be removed. This is a perfectly realistic prospect that takes very little to become reality, so your complacency is unwarranted.

No, we are using the word predictable in the exact same way. I fully agree that ‘a feature that varies in realization or distribution from speaker to speaker isn't something that we can reliably transcribe’. The point is that the devoicing of non-final high vowels is precisely such a feature based on the totality of sources, as each source gives a different description of when non-final high vowels are devoiced. It is against WP:NPOV to pick one source and trust that one while disregarding the sources that disagree with it. We don't get to pretend that the sources that contradict Fujimoto don't exist. We have no basis upon which to determine whether a given word will feature non-final devoicing consistently from speaker to speaker in Standard Japanese, because going by one source, a given word will consistently exhibit devoicing, and going by another source, the same word will consistently exhibit full voicing – and we don't get to decide that one source is right and the other is wrong. What sources say about devoicing of non-final high vowels in Standard Japanese is all over the place (I give three examples in my last comment), so we don't have any basis in the totality of sources for a policy. Deciding to go by Fujimoto and disregard those with a different view would be arbitrary and thus in breach of WP:NPOV. Libhye (talk) 09:10, 2 October 2019 (UTC)

I'm not worried about these editors that you talk about.

Who is contradicting Fujimoto? Certainly not Shaw & Kawahara (2018) or Labrune (2012). — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 19:02, 2 October 2019 (UTC)

Hibiya (1999), for instance. Libhye (talk) 14:22, 3 October 2019 (UTC)

Not sure what that's a reference to, but that's three to one. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 15:27, 3 October 2019 (UTC)

There's also Shibatani (1990), Kawakami (1977) and the acoustic study of Takeda and Kuwabara (1987). The point is that reliable sources contradict one another. Libhye (talk) 15:50, 3 October 2019 (UTC)

Hmmm. Strange that all of the sources you're citing are 20-40 years old while Fujimoto and the other sources are from the last decade. Is there anything more recent that goes along with the narrative of those older sources or has there been a shift in the claims coming from scholarship? — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 16:24, 3 October 2019 (UTC)

The claims coming from scholarship have been shifting back and forth: in 1962, Han said that devoicing occurs in accented vowels; 28 years later, Shibatani said that it doesn't; and now it's another 29 years later. Anyway, this discussion is moot since, as I wrote yesterday, I have now given up building consensus for my original proposal and instead made a new one. So what is your opinion on my new suggestion? Libhye (talk) 11:01, 4 October 2019 (UTC)

Did any of those you mentioned contradict Fujimoto's classification of devoicing environments by frequency specifically? Hibiya (1999) didn't categorically say accented vowels do not devoice, but rightly noted that the accent shift had been disappearing. Shibatani (1990) is just an overview book about languages of Japan, not an empirical study or meta-analysis or even a work about phonetics or phonology in particular. If the point you're making is "Certain claims made by Fujimoto have been contradicted, therefore all of her claims are unreliable", or "Claims about vowel devoicing have been variable, therefore all claims about vowel devoicing are unreliable", that's hasty generalization. Fujimoto does address all of those preceding studies you named, and then comes to a conclusion. We wouldn't be disregarding them by following Fujimoto—rather they are incorporated in her study—unless you can name a newer work that cites and contradicts it. And please, please indicate sources more specifically than just surnames and years. Nardog (talk) 22:40, 4 October 2019 (UTC)

Of course I'm not saying that if some of the claims in an article have been contradicted, then they're all unreliable, or that if claims about X have been variable, then all claims about X are unreliable. I am saying the opposite: that all reliable sources are reliable, even those that contradict Fujimoto. The fact that someone addresses previous claims in an article makes no difference as to the reliability of those previous claims; they do not somehow become obsolete. Nor does it matter which article was published last. If we were to deal with sources the way you imply, then we would react to a debate between two scientists like this: Scientist A publishes an article. Scientist B publishes an article that cites A and contradicts them. Wikipedia concludes that A's study is incorporated in B's study and follows B since there is no newer work contradicting B. Then A publishes a new study citing B and contradicting them. Wikipedia then concludes that B's study is incorporated in A's new study and follows A since there is no newer work contradicting A's new study. But then B produces a second article wherein they address A's second article in detail and contradict it. Wikipedia now goes back to following B's view. Complaints that A's view isn't given due weight are responded to by claiming that A's studies are incorporated in B's latest study and therefore not being disregarded. Then A publishes a third study, and Wikipedia changes its mind once again with the exact same reasoning. And so on and so forth. Libhye (talk) 15:04, 6 October 2019 (UTC)

You keep using the word "reliable" but I don't think it means what you think it means. We mean "reliable" in the sense that it can be used as an authoritative source to back up claims made in an encyclopedia. In that sense, arguing that the sources contradict each other and are therefore unreliable is exactly what you are saying.

In academia, we do, indeed, give priority over more recent sources, all things being equal. It seems that more recent sources lean more heavily towards Fujimoto's claims. It's not convincing to come up with a convoluted scenario of how the situation could be and then expect us to act as if this scenario were true without demonstrating (or even arguing) that this is the case. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 17:28, 6 October 2019 (UTC)

It's a pity you don't seem to understand anything I say. The fact that different studies contradict one another doesn't mean that any of those studies is unreliable. It does mean, however, that we have to give due weight to the different claims. As I have explained in detail, the fact that a source is more recent is in itself meaningless.

I have no idea what you mean by your final sentence. What I have argued we should do in this particular case is mark devoicing wherever it normally occurs. That's a common-sense approach if ever there was one. However, the sources don't tell us where devoicing normally occurs, and so we have a problem. We know from the sources that the lexical aspect is relevant, but Fujimoto and similar sources neglect the lexical aspect altogether and are therefore useless for our purposes: it doesn't help us to know that high short vowels are typically devoiced in a certain position in general; we need to know whether the high short vowel is typically devoiced in the word we are transcribing. That's why my preferred solution is to research each word individually.

If, however, we're not going to transcribe devoicing wherever it normally occurs, it's better to base our decision on when to transcribe it on the practice of pronunciation dictionaries than to base it on articles that say nothing about when devoicing should be transcribed. Libhye (talk) 11:09, 7 October 2019 (UTC)

As I have explained in detail, the fact that a source is more recent is in itself meaningless. I understand your explanation. I don't accept it because, while it could be true in this case, you need to demonstrate that it is true. You haven't and there is therefore no reason to act as though it is true. Without it being demonstrated as true, we would want to give preference to more recent sources that contradict older sources. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 15:20, 7 October 2019 (UTC)

Nardog made a general point, and I showed how his general point was wrong. I have no interest in relating it to my original proposal since, as I wrote four days ago, I have made a new proposal and this part of the discussion is therefore moot. Libhye (talk) 10:00, 8 October 2019 (UTC)

We know from the sources that the lexical aspect is relevant No, you have yet to prove that. Name a study besides Hasegawa (1999) that shows devoicing is stored in the lexicon beyond desu/masu vs. others. Nardog (talk) 18:52, 7 October 2019 (UTC)

There is no need for a study ‘that shows devoicing is stored in the lexicon beyond desu/masu vs. others’. Those two examples alone show that the lexical aspect is relevant. Libhye (talk) 10:00, 8 October 2019 (UTC)

Again, can you name any study (from any period) that contradicts Fujimoto's finding that high-vowel devoicing occurs much more consistently in stop–stop, fricative–stop, and stop–non-/h/ fricative environments than in others? For a topic where there is an actual debate among scholars, like "Is the syllable a useful concept in Japanese phonology?" or "Do /ii/ etc. and /iː/ etc. contrast?", sure, we would be in violation of NPOV if we presented one side as though it's a fact, but when there is no debate taking place, I don't know how one can possibly give undue weight. Nardog (talk) 18:37, 7 October 2019 (UTC)

That's just one of Fujimoto's claims. Many people have written on the subject of vowel devoicing in Standard Japanese, and their portrayals of the overall situation vary (as I've already mentioned, Hasegawa contradicts three previous studies). What Fujimoto and similar studies say is too general to be of use in the transcription of individual words, and what we do know about the pronunciation of individual words is too fragmentary. For instance, when Hasegawa sets out to show Martin is wrong about word-final devoicing being prohibited in accentless words, four of her seven examples of accentless words with final devoicing end with [tsɯ̥], suggesting that devoicing is more common in such words. But it doesn't seem like anyone has done much research into the lexical aspect of vowel devoicing, and so there is a lot that just isn't known. As I wrote above, ‘it doesn't help us to know that high short vowels are typically devoiced in a certain position in general; we need to know whether the high short vowel is typically devoiced in the word we are transcribing. That's why my preferred solution is to research each word individually. If, however, we're not going to transcribe devoicing wherever it normally occurs, it's better to base our decision on when to transcribe it on the practice of pronunciation dictionaries than to base it on articles that say nothing about when devoicing should be transcribed.’ Libhye (talk) 10:00, 8 October 2019 (UTC)

But that's the only claim of Fujimoto's that we (Aeusoes1 and I) are suggesting we follow. So if you're going to dismiss it, then of course we will ask for counterevidence. You are in effect saying that if some of the claims in an article have been contradicted, then they're all unreliable, or that if claims about X have been variable, then all claims about X are unreliable.

One of the studies Fujimoto cites is a corpus study of 300,000 vowels. IMHO that alone should outweigh Hasegawa's seven examples, which are presented with no citation to an empirical study. Fujimoto does address the lexical aspect, and says that while "devoicing between a voiceless consonant and a pause does not consistently occur", devoicing is "very common" in "function words such as desu and masu" in "phrase-final position", which is completely in line with Hasegawa. If she makes no further mention of the lexical aspect in such a comprehensive overview (Shaw & Kawahara's words), then it is only natural to me to think it is of little relevance. Insisting on a lexical condition based solely on the words "with some exceptions" in a sentence in a 1999 conference paper, without even knowing what the exceptions are aside from the two examples given, strikes me as bizarre. If there is a lot that just isn't known, then we make do with what is already known—to do otherwise is WP:CRYSTAL. Nardog (talk) 15:09, 8 October 2019 (UTC)

For the fourth time: I abandoned my original proposal over a week ago. Since then I have not been arguing for it, so your and Aeusoes1's attempts at arguing against it are pointless. I have never dismissed a single claim made by Fujimoto, but you have to admit that Fujimoto doesn't contain a single word of advice on when to transcribe vowel devoicing in Japanese. Your proposal is entirely original research. For the last time: I am not saying – in effect or otherwise – what you insist on misunderstanding me as saying. My point is entirely about giving due weight to different sources, and it is a general point, not one made in defence of my original proposal that I stopped defending eight days ago. WP:CRYSTAL doesn't apply since this part of the discussion is not about what I propose we do; my current proposal is that we follow the dictionaries. That there is a lexical aspect that should be investigated further (by scientists at some point, not by us before we make a decision) is supported by your quotes from Fujimoto. She agrees with Hasegawa that there is something special about a group of words/suffixes containing désu and -másu, and like Hasegawa, she implies that these aren't the only two members of said group. So what Fujimoto writes only strengthens my opinion that the lexical aspect of vowel devoicing merits further research. Unfortunately, such research has not been performed, and so it is hardly surprising that there is not much mention of this aspect in an overview, no matter how comprehensive. What little there is, however, comfirms that the lexical aspect is relevant. Libhye (talk) 18:47, 11 October 2019 (UTC)

@Nardog, Aeusoes1, and IvanScrooge98: It seems clear at this point that it won't be possible to build a consensus for my preferred solution, which is to mark devoicing of high short vowels wherever it occurs. The second best thing, in my opinion, to marking it wherever it occurs, is marking it wherever dictionaries mark it. That would address my concern about us randomly deciding which cases of devoicing to mark. I am not terribly pleased with how dictionaries deal with this issue, as they refrain from marking devoicing in cases where it normally occurs in isolation – and we transcribe words in isolation. But at least basing our decision on the considerations of dictionary editors beats basing it on the musings of Wikipedia editors. – So is there any opposition to this suggestion? Libhye (talk) 14:22, 3 October 2019 (UTC)

@Libhye: no opposition, if that counts (as you can see I stopped intervening because we were getting too deep into something I have little knowledge of). 〜イヴァンスクルージ九十八［IvanScrooge98］（会話） 18:56, 3 October 2019 (UTC)

I don't think anybody would be opposed to citing a source whenever available. But if you're suggesting we only mark it where dictionaries mark it, I'd be opposed to it because, again, devoicing is not recorded in general dictionaries, only in accent dictionaries, which are rather prescriptive in nature as they are reference works aimed primarily at announcers and actors, and don't include many of the words currently transcribed using IPA-ja. I also repeat that I'm opposed to marking it word-finally—because that would falsely suggest it is devoiced even when followed by a voiced sound—except in phrase-final desu/masu. Nardog (talk) 22:40, 4 October 2019 (UTC)

My proposal isn't about citing sources but about when to mark devoicing. I think we should mark devoicing whenever it would be marked in dictionaries that mark devoicing. The alternative to following dictionaries is to invent our own practice. To mark devoicing in some cases where it normally occurs and not in others based on our own judgment would run counter to WP:OR. To mark it in some cases where it normally occurs and not in others based on what dictionaries do would not.

Dictionaries that mark devoicing do so in a gazillion cases, so it makes no difference if a word is not in the dictionaries: it is still clear from the other cases how the word would have been transcribed. I'm surprised, though, at your assertion that we transcribe many words that are not included in pronunciation dictionaries. Would you perhaps name a handful of such words to substantiate your claim? Libhye (talk) 15:04, 6 October 2019 (UTC)

What would be the difference between your proposal and Nardog's? Like, how would we word the explanatory note in the guide. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 17:28, 6 October 2019 (UTC)

The main difference is that my proposal is based on the practice of pronunciation dictionaries, while Nardog's is based on articles that say nothing about when devoicing should be transcribed. My proposal avoids setting the precedence that editors can decide to transcribe a feature only in some of the cases where it normally occurs based on sources that don't say a word about when said feature should be transcribed.

My original proposal was that we should transcribe devoicing wherever it normally occurs, including in ‘désu’, ‘-másu’ and ‘áki’. My secondary proposal to follow the pronunciation dictionaries means that we should not transcribe devoicing word-finally, not even in phrase-final ‘désu’ and ‘-másu’. Libhye (talk) 11:09, 7 October 2019 (UTC)

I already understand your proposal and this is, in no way, an actual answer to my question. Again, in bold, what would the actual wording in our explanatory note be for the two approaches? — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 15:20, 7 October 2019 (UTC)

You wanted to know what the difference would be between my proposal and Nardog's, and I explained that in my proposal, devoicing wouldn't be transcribed in phrase-final ‘désu’ and ‘-másu’, and we would avoid the unfortunate precedence that would be set by Nardog's proposal. How is that not an answer to your question? Discussing how to word the note at this time would be getting ahead of ourselves: first we need to agree on which proposal to go with. But since you ask, the note should probably stay as it is, with a sentence at the end explaining that Wikipedia only transcribes vowel devoicing where pronunciation dictionaries do (or would, when we transcribe words that aren't in the pronunciation dictionaries). Libhye (talk) 10:00, 8 October 2019 (UTC)

My secondary proposal to follow the pronunciation dictionaries means that we should not transcribe devoicing word-finally, not even in phrase-final ‘désu’ and ‘-másu’. Wow, where did that come from? Do the dictionaries prescribe not devoicing desu/masu?! Nardog (talk) 18:43, 7 October 2019 (UTC)

The dictionaries don't transcribe devoicing word-finally. That, of course, does not amount to proscribing such devoicing. Libhye (talk) 10:00, 8 October 2019 (UTC)

As you can imagine, many IPA-ja transcriptions are of the names of things from pop culture (and the names of mountains, for some reason), like Naushika, Pīchi-hime, Pikachū, Pokemon Komasutā, and Sakanakushon. I don't have access to an accent dictionary at the moment, but I doubt these are included. So how do you suggest we ascertain whether the high vowels placed between voiceless consonants in these words are devoiced or not? Nardog (talk) 18:37, 7 October 2019 (UTC)

As I wrote, ‘[d]ictionaries that mark devoicing do so in a gazillion cases, so it makes no difference if a word is not in the dictionaries: it is still clear from the other cases how the word would have been transcribed.’ Libhye (talk) 10:00, 8 October 2019 (UTC)

I'm a bit puzzled by this whole conversation, since my impression is that this unvoiced vowel distinction is very unpredictable. But can you please give me the titles of a couple of these pronunciation dictionaries which I could check? Imaginatorium (talk) 13:37, 8 October 2019 (UTC)

NHK日本語発音アクセント新辞典 and 新明解日本語アクセント辞典 are the two major ones. Nardog (talk) 14:23, 8 October 2019 (UTC)

Which words, though? What word do you suggest we look up when transcribing a word like, say, Pikachū? And wasn't your whole argument that devoicing was lexically conditioned? So let's say we look up whatever the word you suggest, and then how do we know if Pikachū follows the same devoicing pattern as that word? Nardog (talk) 14:23, 8 October 2019 (UTC)

My argument that there is a lexical aspect to devoicing – which is confirmed by Fujimoto – is relevant only to my original proposal, which I abandoned eight days ago as it didn't garner sufficient support. My current proposal is that we follow the practice of pronunciation dictionaries, which don't take into account any lexical aspect of vowel devoicing in Standard Japanese. So we will transcribe Pikachū with a voiceless i, just like any other word with an i between a p and a k, such as píkata. Libhye (talk) 18:47, 11 October 2019 (UTC)

I still prefer Nardog's proposal that we should mark devoicing only in the environments that Fujimoto designates "typical". — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 19:01, 11 October 2019 (UTC)

Nardog's proposal is to mark devoicing not only in those environments but also in phrase-final désu and -másu. My proposal doesn't lead to a very different outcome than his proposal; the most important difference is that my proposal is based on an established practice while his proposal is completely without precedence. For the examples that Fujimoto labels ‘moderately frequent’, NHK日本語発音アクセント新辞典 (2016) does not mark devoicing in sushi and shihei but does so in shikki. Nardog's proposal would ban editors from marking devoicing in many cases where pronunciation dictionaries do, while my proposal, much more practically, would allow them to simply copy the dictionaries. Libhye (talk) 21:03, 11 October 2019 (UTC)

Nardog's proposal would ban editors from marking devoicing in many cases where pronunciation dictionaries do I never said such a thing. My proposal is to mark devoicing in environments Fujimoto designates as typical by default. It doesn't preclude citing and following an accent dictionary for words for which one is available.

Our transcriptions are not bound by how dictionaries transcribe words, let alone those which do not use IPA or even an alphabet. Accent dictionaries do not transcribe word-final devoicing because it is obligatory only in function words and its users are expected to be proficient in the language enough to have it already acquired or at least be cognizant of it. If reliable descriptive sources (including the dictionaries themselves) say desu/masu are regularly devoiced phrase-finally, there is little reason not to mark them as such.

The problem in your proposal is that accent dictionaries do take lexical aspects into account. The NHK dictionary explicitly notes this in the preface (p. 16): この辞典では、このような条件を基本におきながら、アクセントの位置や語の複合の構造、無声化が複数続く場合などを考慮し、語ごとに判断して無声化記号を付けた。 ("With these conditions [high vowels between voiceless consonants] as the basis, this dictionary marks devoicing judging word by word, taking accent position, compound structure, consecutive devoicing, etc. into consideration.") So following accent dictionaries for words which they do not transcribe will inevitably involve guesswork, and therefore is OR. Following Fujimoto is not. Nardog (talk) 08:35, 13 October 2019 (UTC)

Sorry, but there's no way that doesn't count as a change of your proposal. Aeusoes1 interpreted it the same way: ‘that we should mark devoicing only in the environments that Fujimoto designates "typical"’. But I appreciate that you've changed it.

The problem with transcribing phrase-final devoicing only in désu and -másu is that it lacks support in the sources. Both Fujimoto and Hasegawa say that désu and -másu are merely examples of that type of words/suffixes. We have no basis for singling out désu and -másu. There would be a basis for transcribing phrase-final devoicing in the entire group, but neither Fujimoto nor Hasegawa tells us what the other members are.

There is no possibility of ‘following Fujimoto’, for unlike the pronunciation dictionaries, she doesn't say a word about when vowel devoicing should be transcribed. Fujimoto only points out some very general tendencies, and working out the transcription of individual words based on that would involve a higher degree of original research than working it out based on pronunciation dictionaries' transcriptions of parallel words. As you're now finally acknowledging the lexical aspect of vowel devoicing – another change in your stance I appreciate – there's no longer a basis for implementing a general rule. Creating a default practice based on what amounts to mere tendencies makes no sense if we all agree that vowel devoicing is lexically determined. Libhye (talk) 20:06, 14 October 2019 (UTC)

This is goofy reasoning. Whether Fujimoto recommends something for transcription is immaterial. Just because Fujimoto is being descriptive doesn't mean that it's impossible to follow that description as a guideline for our in-house practices. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 20:26, 14 October 2019 (UTC)

@Aeusoes1: I haven't said it is impossible to follow Fujimoto because she is descriptive. But given that we transcribe individual words, it makes more sense to derive our transcriptions from the more detailed practice of pronunciation dictionaries than from the overly general rule that Nardog is proposing. Would it be more acceptable to you if I actually made an explicit proposal based on the practice of pronunciation dictionaries, or do you somehow disapprove of their practice? Libhye (talk) 23:29, 20 October 2019 (UTC)

Is this a joke? From your above comment: there is no possibility of ‘following Fujimoto’ Well, saying something has no possibility sure sounds like you're saying it's impossible. As to the reason? You say it's impossible because Fujimoto only points out some very general tendencies that is, she is taking part in descriptive linguistics, rather than taking on prescriptive declarations you find in dictionaries that say things like when vowel devoicing should be transcribed. I mean, do you really expect me to believe that this means something other than the plain reading I identified? That's a bit insulting.

I've already asked twice for actual wording behind your proposal and you've twice declined. I'm currently against your proposal, but (if we're assuming good faith) you've so far demonstrated such poor writing in the talk page that Nardog and I have several times been told that our plain reading of your comments has been incorrect. Now I'm wondering if what you'd like to see isn't something different from what we think you'd like to see based on your comments. So if you're so disposed, I ask (a third time) for your proposed wording. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 03:05, 21 October 2019 (UTC)

Like I said, I don't think anybody would be opposed to citing a source whenever available, so I just assumed that was presupposed in our conversation. I didn't change my stance, only clarified it.

Your second paragraph doesn't make a lick of sense to me. If we find out what the other members are, we'll just add them to the list of words we always mark as devoiced.

I'm in no way "acknowledging the lexical aspect of vowel devoicing". I'm just saying accent dictionaries do. I'm skeptical it is, or can be, empirically substantiated because, in Fujimoto's words, Contrary to the description that devoicing is obligatory in Tokyo Japanese, its actual occurrence diminishes due to many factors such as consonantal environment, accent, speech rate, and dialects (167). And, again, since not all words we transcribe are included in the dictionaries, we absolutely need a general rule so long as we mark devoicing at all.

And what Aeusoes1 said. None of our language-specific IPA transcription schemes stricly follow an existing set of conventions anyway. Nardog (talk) 06:34, 16 October 2019 (UTC)

@Nardog: In my second paragraph, the point is that it would be misleading to transcribe only phrase-final désu and -másu with a voiceless final vowel if a greater number of words/suffixes behaves exactly like them, since that would imply a relevant difference between désu and -másu on one hand and the rest of the group on the other. No such relevant difference exists, so there is no justification for treating désu and -másu differently from the rest of the group.

We have to follow what the sources say, no matter what you personally believe. Do you have a source that says there is no lexical aspect to vowel devoicing in Standard Japanese? Fujimoto indicates that such a lexical aspect exists, so you can't use something she says as an argument against it.

When transcribing words that aren't in accent dictionaries, transcribers have to research the pitch accent information themselves. If they can be trusted to do that themselves, then they can also be trusted to research the vowel devoicing themselves. There is therefore no actual need for a general rule. Conversely, if we aren't to trust them on pitch accent, then there is no reason to trust them to transcribe such words correctly at all. The logical consequence of your insistence on a general rule is that we can't transcribe words that aren't in pronunciation dictionaries. If that is not your stance, then please explain how your stance makes sense. Libhye (talk) 23:29, 20 October 2019 (UTC)

Note for [ɯ]

@IPA editor: I'm not disputing your addition because I think it's inaccurate. It's because I find it too insignificant to include in a guide for layreaders. Also note that Okada (1991/99) is merely talking about Tokyo speech. Especially in and around Osaka, /u/ is distinctly back and rounded. At the other end of the spectrum, it can be as front as [ʏ]. There's just too much variation. This is why the note avoids going into phonetic details. And since [ʏ], [ʉ], [ɨᵝ] and [ɯ] are all acoustically very similar, [ɯ] does the job of giving the readers a target perfectly adequately. Nardog (talk) 14:34, 10 January 2019 (UTC)

OK, got it. IPA editor (talk) 17:15, 10 January 2019 (UTC)

You forget some allophone of japanese /u/ sound. It's “[ÿ]”. Juidzi (talk) 08:52, 21 April 2020 (UTC)

That's the same as [ɨᵝ]. (By the way, we should completely abandon the practice of using the symbols for front rounded vowels to signify compressed lips. It has little if no precedent in literature, and the canonical lip configuration of those symbols as conceived by Daniel Jones is protruded.) Nardog (talk) 09:24, 21 April 2020 (UTC)

About “[ç]”

Why I can hear the “[x(ʲ)]”? Please! explain it. Juidzi (talk) 06:17, 23 April 2020 (UTC)

Your question is almost unintelligible, but if you're asking why [ç] sounds like [x] or [xʲ], well, of course it does because they share the manner of articulation and are pronounced at very close places of articulation, hard and soft palate, and ⟨ʲ⟩ means the tongue is raised towards the hard palate. Listen to and look at the videos demonstrating [ç] and [x] here and here. Nardog (talk) 12:42, 23 April 2020 (UTC)

Also, in some transcription systems ⟨xʲ⟩ can actually stand for [ç] as there may be phonological reasons to transcribe it as such (and vice versa - ⟨ç⟩ can stand for a post-palatal fricative, rather than a true palatal. This is the case in Greek). Kbb2 (ex. Mr KEBAB) (talk) 13:14, 23 April 2020 (UTC)

Jeep

Why is the English approximation for /dʑ/ jeep when everybody pronounces Jeep as /dʒip/. Nobody pronounces Jeep as /dʑip/. — Preceding unsigned comment added by Stunts1990 (talk • contribs) 13:25, 20 May 2020 (UTC)

Because it's an approximation and [dʑ] is not found in any standard variety of English. Nardog (talk) 13:38, 20 May 2020 (UTC)

"English" approximations

For vowel お (o) there is "story". I am rhotic, from SW England, and there is no similarity between the Japanese vowel and my pronunciation of "story". So I added "American"; I'm not sure that even this is true. But in any event it would surely be more helpful to have separate sets of approximations for AmE and BrE...? Imaginatorium (talk) 18:45, 9 August 2020 (UTC)

It is close to Japanese /o/ in both GenAm and RP. Although typically transcribed with ⟨ɔ⟩, the modern realization is between cardinal [o] and [ɔ] in both accents. GOAT is clearly inferior because it's diphthongal and unrounded/centralized in both, especially RP.

I know not all accents of English share these characteristics but they are too numerous to account for and RP and GA are what non-native speakers, who must make up a substantial portion of our readership, are familiar with and what are usually meant by "English" when there is no qualification, even in scholarly works. The keys for Spanish and Hebrew, which have similar five-vowel systems, also use NORTH/FORCE to illustrate mid [o̞] and no one has complained. I see no compelling reason to change it or add a qualification. Nardog (talk) 03:51, 10 August 2020 (UTC)

Are you claiming that if I say "sorry" and "story", the vowels are the same? They aren't. I don't quite know what else to say. The 'o' in "story" is a long deep "aw" sound, like "raw" or "wrought". These are wildly different from "rot", which is indeed like Japanese 'o'. What version of English do you speak, and have I inadvertently switched planet? OK: dictionary, LDEL, respelling: 'sorry' /sori/; 'story' /stawri/. I think what you want for Japanese /o/ is the respell "o", not "aw". But I understand that the American pronunciation of "sorry" is closer to Japanese /a/, which is why lots of loans from English are transcribed with /a/. So I am forced to say "callege" instead of "college". Imaginatorium (talk) 04:28, 10 August 2020 (UTC)

Oh, so you're saying LOT is closer to Japanese /o/. Though it might be in your accent, it is in neither GA nor RP. In GA it is low, unrounded and centered, as you pointed out, and in RP it is still too low compared to Japanese /o/. And since length is not by itself contrastive in most accents of English, the qualitative proximity of the NORTH vowel makes it the optimal approximation (which is why a polysyllabic word like story is preferable to a monosyllable like more and which is why we use father for central [ä] in most of the keys for languages with such a vowel, even though PALM is technically a tense vowel). Nardog (talk) 04:44, 10 August 2020 (UTC)

I am never too sure what "RP" means; I have to try to imagine listening to a film from the 1940s. I confused the situation by talking about being rhotic (just means I distinguish 'fought' and 'fort'), but I basically have a standard southern England "educated" pronunciation. Exactly like Victoria Coren Mitchell, for example, except that she does not distinguish 'fought' and 'fort'. Can you give some sort of reference that suggests why I might think that LDEL's respelled 'aw' is more like お than its respelled 'o'? Imaginatorium (talk) 05:06, 10 August 2020 (UTC)

Just compare the vowel quadrilaterals shown in Japanese phonology#Vowels and Received Pronunciation#Vowels. Notice /ɔː/ is placed far closer to Japanese /o/ than /ɒ/ is. The RP diagram is based on Roach (2004), as the caption says. There aren't many native speakers of RP anymore, granted, but it's still the accent most frequently used as the reference to compare other accents to, and as the model for learners of English to acquire. Nardog (talk) 05:16, 10 August 2020 (UTC)

Cruttenden (2014) and Lindsey (2019) note that LOT is open-mid in contemporary RP, so it's a very good fit for the Japanese vowel. But, as you say, the American pronunciation of LOT is the opposite of a good fit. Sol505000 (talk) 15:24, 10 December 2021 (UTC)

[ɔ] would be too low. English speakers often substitute /u:/ for /o/ when repeating a Japanese word they just heard. /u/ being central if not front (Maekawa 2021 found it's more front than /a/), the high back region is vacant in Japanese, which I suspect contributes to /o/ being closer to [o~u] in perception and/or production. Nardog (talk) 11:53, 1 January 2022 (UTC)

Palatalization

This chart shows the palatalized sounds separately, but some of the unpalatalized examples are before /i/, which causes palatalization. If it were for unpredictable palatalization, there wouldn't be examples of palatalized sounds before /i/ since it's predictable there. Why isn't it directly phonetic? ThighFish (talk) 18:37, 14 October 2021 (UTC)

See /Archive 2#き as kʲ is inaccurate. It's that way because some people thought showing ⟨ʲ⟩ before [i] was too complicated. I for one can see either way: showing it would indeed be more consistent as you say, but it's also a coarticulation effect you find in about every language. Nardog (talk) 12:02, 1 January 2022 (UTC)

Utterance-final nasal

@Imaginatorium: You seem to have confused /N/, which is an archiphoneme, with [ɴ], which is a phone. If you accept that it is not a lie to represent the utterance-final nasal with ⟨ɴ⟩, then you have to transcribe all syllable-final nasal with ⟨ɴ⟩ instead of ⟨m, n, ɲ, ŋ, ɰ̃⟩, because now you're using an archiphoneme, and tell the readers that it doesn't represent what it represents on the IPA chart. But we use phonetic transcription for all non-English languages, as indicated by the square brackets as opposed to slashes.

@Aeusoes1: What I don't understand about your edit is why you removed this note in particular. We have notes about [w, ɰ̃, ɯ], which are also more convenient conventions than are accurate transcriptions, and I don't see how they're less "unnecessary for understanding transcriptions or transcription conventions" than the one about ⟨ɴ⟩. Nardog (talk) 09:26, 12 February 2022 (UTC)

The footnotes should be designed to help readers understand our transcriptions and for editors to understand our transcription conventions. The three notes you've identified do the latter, [ɰ̃] most plainly so. With [w] and [ɯ], the notes providing detailed phonetic information due to a tendency for occasional edit wars over transcription conventions due to the perception of phonetic inaccuracy. I hadn't seen the same need for this with ⟨ɴ⟩, but given Imaginatorium's recent edits, I can be convinced otherwise. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 18:10, 12 February 2022 (UTC)

My concern is that ⟨ɴ⟩ in our transcriptions doesn't mean what it means on the IPA chart, and readers should know about it. We already state it at Japanese phonology and Voiced uvular nasal, so anyone who's read them will inevitably have "the perception of phonetic inaccuracy". And yes, if it was supposed to be, or could be mistaken for, an archiphoneme, that's all the more reason to tell the readers about it. Imaginatorium's summary is therefore not only a poor rationale for removing the note but a compelling one for keeping it, ironically. Nardog (talk) 18:55, 12 February 2022 (UTC)

Fair enough. Perhaps a reword that gets at the heart of the matter might be in order. Something like The actual realization of the utterance-final nasal varies; ⟨ɴ⟩ is a generally accepted convention based on the common belief that it is realized most often as a uvular nasal. (Maekawa 2021) — Preceding unsigned comment added by Aeusoes1 (talk • contribs) 04:30, 13 February 2022 (UTC)

Why has 'わ' been written as w as in wasabi in the table when わ is pronounced as ne?

わ in wasabi(romanisation) isn't it incorrect? 103.15.254.86 (talk) 08:28, 3 May 2022 (UTC)

I assume you've confused わ with ね. Nardog (talk) 08:57, 3 May 2022 (UTC)

The page seems to ignore the use of [ŋ] sound (鼻濁音)

I'm a random native Japanese speaker and not an expert in phonology, but I suppose the article ignores the existence of [ŋ] sound (鼻濁音) in Japanese. The sound is often used instead of [g] in the middle or at the end of a word, particularly in the normative way of pronunciation. Thus it seems to be inappropriate to refer to "きぎょう (企業 or 起業)" as the example of [ɡʲ] sound on the table. Some people (such as newscasters) use [ŋ] and pronounce the word as [kiŋʲoː] instead of [kiɡʲoː]. An appropriate example is "ぎょうせい (行政)", which is 100% certainly pronounced [ɡʲoːseː] because the sound in question is at the beginning of the word. You can refer to ja:鼻濁音 if necessary. Tooooriiiiii (talk) 13:06, 27 December 2022 (UTC)

Unlike Japanese phonology, this is not an article but a guide for readers on how to figure out what each symbol in transcriptions in articles means, and for editors on how to transcribe Japanese words. And it doesn't "ignore" [ŋ] as an allophone of /ɡ/—see note 3. It just transcribes all instances of /ɡ/ with ⟨ɡ⟩ because [ŋ] is a regressive allophonic feature confined to certain regions and generations. Vance (2008: 214), which the note cites, says: According to descriptions written around 1940, many Tokyo natives were consistent nasal speakers at that time, but the proportion of inconsistent speakers and consistent stop speakers was on the increase. More recent sociolinguistic studies indicate that the proportion of consistent stop speakers has in fact increased and that, at least in some Tokyo neighborhoods, the proportion of tokens pronounced with [ŋ]/[ŋʲ] correlates with the age of the speaker. People may throw it in every now and then especially in formal registers, but very few people now consistently have [ŋ] where it's a possibility. Nardog (talk) 15:27, 27 December 2022 (UTC)

Thank you for your prompt reply. That makes sense. I was overlooking the notes. Tooooriiiiii (talk) 08:53, 29 December 2022 (UTC)