Pattern to H-mutation (into P or B)?

Sorry if this is obvious to anyone but I just failed a vocab review by putting げんぼう when it should have been げんぽう and it got me thinking about H-mutation in general.

The H-row sounds は・ひ・ふ・へ・ほ often turn into either ば・び・ぶ・べ・ぼ or ぱ・ぴ・ぷ・ぺ・ぽ when they get compounded. Fair enough. Are there broad and useful tendencies to these changes? I thought about it and did some research, and it seems:

  • kun’yomi compounds seem to favour H turning into B (yay rendaku)
  • on’yomi compounds seem to favour H turning into P after ん (counter words are a big exception)
  • H turning into P happens after gemination (doubling) regardless of kun’yomi or on’yomi

Kun’yomi compound examples would be 落ち葉 (おちば) and 頑張る (がんばる).
Jukugo (on’yomi compound) examples would be 文法 (ぶんぽう) and 電波 (でんぱ).
Examples of gemination (doubling) during compounding would be 引っ張る・ひっぱり (引き + 張る) (kun’yomi) and 失敗・しっぱい (jukugo).

This would imply the following heuristics:

  1. In jukugo or kun’yomi compounds, (っ)ぱ could very well have started life as a は which got geminated/doubled (き/ち/つ+は geminates to っぱ across morpheme boundaries, though く+は doesn’t seem to trigger gemination except in counter words… ?)
  2. Aside from gemination, ぱ in a kun’yomi compound is not mutated from は because non-geminating sound changes in kun’yomi compounding are all about the rendaku (so ば and not ぱ)
  3. ば in a kun’yomi compound may well have been rendakued from は - e.g. 落ち葉・おちば
  4. In a jukugo word, a reading beginning with B (e.g. ぶん) in a jukugo (e.g. 新聞・しんぶん) is almost never mutated from an H ふん with the major exception again being counter words, e.g. 三分・さんぶん
  5. In a jukugo word, a P-initial reading after ん (e.g. ~んぷん) comes from an H/F-initial reading (e.g. ふん)… with two exceptions: the country kanji for Portugal and Poland

These seem to be broadly if not absolutely on the money. Anyone care to shoot my procrastination full of holes? (Please do!)

7 Likes

I’m sure there are general rules to rendaku, but they are either too many to memorise, or have too many exceptions, or both.

Example:

えん+ひつ=えんぴつ
まんねん+ひつ=まんねんひつ (although まんねんぴつ and まんねんふで are also listed as alternative readings for 万年筆)

1 Like

Good catch! Thank you!

H going to P shouldn’t be rendaku though - p isn’t voiced.

1 Like

Yeah I was using the term rendaku liberally.

三杯=さんばい
乾杯=かんぱい

1 Like

It’s funny how numbers are a major source of exceptions! I couldn’t find examples of jukugo h- going to b- after n outside of number words, seems to be almost exclusively a kun’yomi thing. Same for specifically -ku + h- becoming -pp- aside from numbers, e.g. ろく + ひき realising as ろっぴき for “six animals”.

Also, 三分 is read differently depending on whether your counting parts (さんぶん) or minutes (さんぷん)

Similarly with 五分 which is ごぶん for parts and ごふん for minutes.

EDIT: I guess this is mostly a consequence of 分 itself being read ぶん for parts and ふん for minutes though…

2 Likes

Also 十分 (じゅうぶん) means enough (this is a word that has many English counterparts)
十分 (じゅっぷん) means ten minutes

Waiting for Leebo to post

7 Likes

I remember reading that long ago, there was no distinction between the は-row sounds and the わ-row ones (which is a reason you have こんにちは the way it is, and also why the kanji 原 can be read either はら or わら, among other quirks). Apparently in old Japanese, all the は-sounds were actually pronounced with a P consonant. Unfortunately I don’t remember where I read this, but maybe someone more linguistics-minded like @Carvs has more information.

My point is, I’m not really sure that the ば or ぱ sounds are transformed は sounds at all, my hunch is that they may have evolved in parallel somehow from the same origin (what if instead of H becoming P, the phenomenon was the other way around?) . Also, the voiced markers (゛and ゜) signs on hiragana weren’t even a thing in writing until the spelling reforms of the 20th century.

I’ll try to search around some more when I get home, this is an extremely interesting topic for me.

In the meanwhile, there’s the Wikipedia article on kana usage (https://en.wikipedia.org/wiki/Kanazukai) that has a little info on the history of kana orthography.

3 Likes

This was also discussed in this thread. :relaxed: We even found some exceptions there, so have a look. :open_mouth:

4 Likes

Indeedy. Taking 分 ふん as a general example of on’yomi morphemes beginning with h-, the ふん reading becoming ぷん from doubling or following ん would be a phonetic rule. There’s a chart on Wikipedia which seems to indicate ふ changes differently to the other mora on the h row so this specific example might be part of the confusion!

分 ぶん just acts like ぶん and it’s not uncommon for on’yomi to begin with “b” - on’yomi starting with “p” appear to be very rare though.

Indeedy! I’ve read this myself on JSX and around the place - /*p/ lenited into the consonant at the beginning of ふ (even when in front of は etc) then lenited even further to /h/, though in ひ it’s more palatalised. So yes, historically that seems to be how things have worked.

Edit: Wikipedia says by Late Middle Japanese, the original /p/ had been changed into /h/ but they started using the /p/ sound again for a separate phoneme. Huh. No wonder this is so confusing…

People generally don’t know the history of their own language though, so it could be that nowadays there’s an underlying sense of a phonetic rule which takes a modern /h/ and turns it into something else under certain conditions - it reconstructs the effect but from the opposite direction.

One way to tell might be to give a native Japanese speaker a counter which begins with h but doesn’t exist yet (e.g. ほい), then ask them to count to ten with that counter using いち, に, さん, よん and see what happens. If they say いっぽい instead of いちほい, the rule is understood phonetically.

2 Likes

Leebo is on vacation.

20 Likes

I’ve read a lot through articles descriping rendakus, havent found any real rules, only a lot of hints.
My impression of redaku is a thing of words transforming over time to support easier pronunciation. Say the word 作家(さっか) from what I think probably started out as さくか which is is a little annoying to say when speaking quickly. So poeple just started saying さっか.
therefore my impression is, whenever a compound pairs two phones that hard to speak aloud in quick succension, it gets rendkued to somethign easier. Since I figured this out, I am able to guess about 80% of the rendakus before WaniKani tells me. That being said, I studied the human speech apparature more than the average person due do doing a lot of speech recognition stuff in university. So i kinda know what sounds are made where and how they’d sound when disformed to compliment another sound.
Speaking the word out loud might help. If it feels unnatural, there might be a rendaku.

1 Like

さっか doesn’t contain rendaku, just to be clear though.

3 Likes

I have been summoned, and I bring some information… I hope!

That’s disputed. Basically, it’s hard to date the change so people have different opinions. From what I’ve seen, I’m partial to Frellesvig’s account, which I talked a bit about in the other post. But basically, in A History of the Japanese Language, he argues in favour /p/ being retained in Old and Early Middle Japanese, with a first change to /w/ in intervocalic position, and then from /p/ to /f/ (more or less) for the remaining /p/ in Late Middle Japanese only. In that version, /p/ didn’t get reintroduced in LMJ but rather it stayed as an available sound and new uses were coined then. The view exposed on Wikipedia is, I believe, the traditional one, more or less, that /p/ had already become /f/ by the classical period and either then or later became /w/ in medial positions; that’s the stance of, say, Vovin, in A Reference Grammar of Classical Japanese Prose, if I remember/understood correctly. Can’t say I’m deeply invested in the topic, though, so that’s about as much as I know. :woman_shrugging:

Well, that much is certainly true… in fact, I know more about the history of the Japanese language than the history of French (though I probably know more about the history of French than the average French person)… doesn’t help that my Latin is shamefully almost nonexistent. But they didn’t have anime in Latin… /o\

Anyway, there are no hard rules AFAIK, but I can comment on a couple of things you mentioned, if you want… no guarantees, though, my phonology knowledge is fairly limited. As they say: “provided as is”.

This would be related to the no-singleton-P constraint: you can’t have a /p/ unless after /Q/ (geminated) or /N/ (post-nasal). This would seem to hold for “non-foreign” words only, though, since in Western borrowings, there are plenty of examples.

Well, that’s kind of (mostly?) true, but due to two factors:

  • As mentioned above, both the native Japanese (“kunyomi”) and Sino-Japanese lexica respect the no-singleton-P constraint, so that leaves you with only /Qp/ and /Np/.
  • Now, there is another constraint that converts /Np/ to /Nb/; Frellesvig says this used to be universal in EMJ, but was lost in LMJ, but even then it mostly only resulted in Sino-Japanese formations containing /Np/ and I can’t find any surviving examples of native words that contain /Np/. I have seen it listed as a rule elsewhere that /Np/ does not occur in synchronically (modern) native vocabulary at all.

So all that remains for native vocabulary (with possibly very rare exceptions), hence kunyomi, is /Qp/.

This I am (even) less sure, but I would guess this is because Middle Chinese had no /h/ so everything was imported as /p/ or /b/, and it only became /h/ through the overall process that mutated /p/ to /f/ to /h/… but only in certain kanji, assuming the words existed when the change took place and participated in it (weren’t analysed as transparent compounds), and had been assimilated as full words by the time the normative go-on/kan-on readings were established. That’s a lot of ifs…

Overall, if you asked a native in modern times how they would pronounce imaginary compounds, I think the result would depend largely on how they felt about the words.

If you tell them they are loanwords from English, they’ll surely try to keep the pronunciation as close to the original as possible; say you tell them ほい is from English “hoy” (with whatever credible meaning you can invent), I believe they’d produce さんほい.

If on the contrary, you tell them it’s a long lost old Japanese counter from the Nara period, even though it may be all lies, if they believe you, they will probably apply all the good old rules and give you さんぼい.

6 Likes

My bad, thought the shortening was also called rendaku. Thanks for clearification.

Yeah, that phenomenon is called gemination.

1 Like

Thanks, this is exactly the kind of explanation I was hoping for. I find historical sound changes a fascinating topic.

1 Like

Wow! Thanks everyone for posting, but especially…

Thank you so much! :slight_smile: (BTW would Frellesvig be worth picking up for someone with an undergraduate degree in linguistics. It’d be for general interest and also to mine it for material to help with Japanese learning, e.g. the suru/aru thing.)

Still, no wonder it’s so confusing. :slight_smile:

I guess my main reason for starting this topic was to have a sieve for “reconstructing” compound readings from kanji, e.g. げんぼう for 減法 - as long as I know 減 is げん and 法 is ほう, I want to be able to discard げんぼう as a possibility because ほう only reads as ぽう during gemination (Q) or post-ん (N) and never ever as ぼう. Specifically for checking my mistakes, that kind of thing.

It’s a bit antithetical to good science but there’s something to be said when a language learner can treat a predictable diachronic sound change as a synchronic one in circumstances like this (inter-morpheme post-N/Q fortition of /h/ to /p/) as long as it’s understood to be a fictitious convenience instead of an academically rigorous account of how the changes actually happened.

This “no-singleton-P constraint” you mention…

…is jolly good to know! Mimetic words seem to get a free pass on where P appears too.

Again I guess I’m more interested in a sieve to catch bad guesses than a good account of what actually happened; I think the いっぽい/いちほい thing was me getting a bit off-track. :slight_smile:

Thanks for your excellent input, Carvs!

That would also be my guess - ease of pronunciation is a major driver for this kind of compounding-driven sound change. But it has to be practical. Since き, く, ち and つ all become っ/Q in compounds, there would have to be a trivial amount of information lost in making them all sound the same, otherwise it would be inconvenient and people wouldn’t do it.

2 Likes

Indeed, mimetics are also a separate category of words, quite interesting in themselves, and with their own rules, cf. Hamano which I discussed in another thread.

I’d say it’s a good read regardless of one’s background in linguistics or not. Pretty accessible and very interesting. Is perhaps more opiniated compared to other references, which you might consider a good or bad thing.

Personally, I keep at hand three main references for historical / classical Japanese: Frellesvig, Vovin, and a copy of McCullough’s Bungo Manual, which is my favourite short reference for morphology. Frellesvig is great for Old Japanese, most notably, but it (justifiably) lacks a systematic coverage of all the various auxiliaries and such found in the classics, which Vovin deals with more thoroughly.

The aru/suru support question is very interesting (and useful for learners) but I don’t know of any single resource that covers it all in one place in a focused fashion, though the info is certainly there, spread over various works.

2 Likes