[prewett]

My understanding of the history of Chinese writing is that it kept trying to go phonetic, but each time they prevented it, because the writing had to be read across an empire with multiple languages. Even so, something like 20% of the characters are quasi-phonetic, with the radical giving the topic and the rest of the character giving the approximate pronunciation, so "the word for {a plant|a thing of metal|a person|etc} that is sounds similar to X".

When the Japanese imported it, they used the characters much more phonetically. They used the whole word when that worked, but the characters got assigned to the Japanese pronunciation of the word, as well as the pronunciation from the pieces of other words where that character appeared, as well as the Chinese pronunciations. Then six hundred years or so later they imported them again, by which point the characters had evolved in Chinese but not in Japanese. So its sort of phonetic, but it's a complete mess.

[kragen]

That's mostly right.

Not 20%, more like 90%. https://en.wikipedia.org/wiki/Chinese_character_classificati...

In Japanese at this point most kanji have an onyomi (the sound of the Chinese word, which has been adopted into Chinese the way Latin words like "adopt" are adopted into English) and at least one kunyomi (the sound of a synonymous Japanese word not derived from Chinese). This does add difficulty but it is somewhat compensated for by the smaller repertoire of characters used in Japanese. A lot of the most common Japanese words, all loanwords from languages like English, and all the inflectional suffixes are normally written with one of two purely phonetic syllabaries.

[thaumasiotes]

I would take issue with

>> When the Japanese imported it, they used the characters much more phonetically.

Japanese kanji are much less phonetic than Chinese hanzi. For hanzi, you can ask "how is this character read?", and it's a simple question with a simple answer, because that question is the basis of the writing system. Kanji are assigned all kinds of different readings on the theory that what really counts is the semantics.

For example...

>> They used the whole word when that worked

Not even in the oldest Chinese writings do you see one character representing a multisyllabic word. Identifying characters with words rather than syllables is an innovation on the part of the Japanese.

Tangentially, you mentioned that the vast majority of characters are phono-semantic compounds. I've been watching some youtube videos in which Japanese people are presented with kanji of varying levels of obscurity and asked to speculate on their pronunciation. Without fail, when they don't know the answer, the interviewees speculate that the two major components of the character both contribute to its meaning.

And that always surprises me because a two-meaningful-components construction is so rare in the character system. Almost all characters aren't constructed from two meaningful elements, and I would have thought the Japanese would be familiar with that fact even though they can't understand the phonetic hints. Do you think this is more of a case of them not knowing how characters are formed ("ignorance"), or more of a case of them speculating on the meaning of each component purely because they don't have the ability to speculate about the phonetics ("searching under the lamppost")?

[Particularly where the obscure kanji are part of an obscure phrase borrowed from Chinese, speculating about the phonetics would be helpful to the problem, but I'm assuming most Japanese just plain don't know what kinds of sounds a Chinese phonetic component might be hinting at.]

[kragen]

I think mostly you're talking about kunyomi. Some do have multiple onyomi (大 being the most obvious example) but it's not common, and it's not a case of "all kinds of different readings on the theory that what really counts is the semantics".

I agree with your "searching under the lamppost" theory. If I'm Japanese and I see an obscure kanji, its phonetic (component) only gives me information about its onyomi, which is almost certainly some wildly obscure loanword from late medieval Chinese that isn't even in my recognition vocabulary, much less my productive vocabulary.

It might also be true that I don't know offhand the onyomi of other characters with the same phonetic—in real life I'm a native English speaker, a second-language speaker of two or three Romance languages, and the kind of person who likes to go around thinking about obscure etymological trivia, but I was today days old when it first occurred to me that "suspicious" was probably etymologically "overseeish", cognate with "perspicuous", "spectrum", "speculum", etc. (I was right about "spicious" but wrong about "sus": it's actually "uplookish".)

So it would be totally unsurprising for even a highly literate native Japanese speaker to know the onyomi of numerous characters sharing the same phonetic, but not have that shared sound come to mind when looking at a novel character with the same phonetic, even if they can guess which part is the phonetic and that the obscure word is in fact Chinese in origin. And the YouTube video is likely edited to focus on the people who got things most entertainingly wrong.

[thaumasiotes]

Yes, I'm talking about kunyomi; I don't see how that affects the point.

> If I'm Japanese and I see an obscure kanji, its phonetic (component) only gives me information about its onyomi, which is almost certainly some wildly obscure loanword from late medieval Chinese that isn't even in my recognition vocabulary

I was highly amused when one of the "obscure" kanji presented was 論. It doesn't get much less obscure than that.

On the other hand, none of the interviewees recognized it. All of them got 檸檬. (Another one I wouldn't have expected to be obscure, but in that case I would have been right.)

It's possible that there's some selection for interviewees who provide entertaining responses, but the effect seems weak to me for two reasons:

(1) Of the people featured in a video, you see all of them respond to every prompt. So selection has to be limited to "which people are going to appear", and even then if an entertaining person is boring on some prompts you'll still see that.

(2) Getting things wrong almost always consists of making one or two wrong guesses and saying "I don't know". It's much more entertaining when people get them right, unless they turn out to be obvious and everyone gets them right.

> (I was right about "spicious" but wrong about "sus": it's actually "uplookish".)

You'd need super, not sub, for "up" or "over". Traditionally it's "over"; I'm not sure if there's another Latin prefix that gets used for "up". There is one for "down", de (as in "depend", "descend", or even "defenestrate"), but nothing comes to mind for "up". The opposite of "descend" is "ascend", where the prefix just means "toward".

[kragen]

Presumably when prewett said "When the Japanese imported it, they used the characters much more phonetically," they meant onyomi. (Perhaps they didn't know about kunyomi.) When they say, "but the characters got assigned to the Japanese pronunciation of the word," that is mostly wrong as a description of how Japanese writing uses kanji today, but I think that is the process from which kana came: https://en.m.wikipedia.org/wiki/Man%27y%C5%8Dgana and apparently it isn't quite extinct: https://en.m.wikipedia.org/wiki/Ateji

There are kanji less obscure than 論. There are some that are so common that even I know them. It is in the 6th grade jōyō kanji, but I doubt it's in the top 1000 most common characters in Japanese, maybe not even the top 2000. You could probably guess that 侖 is the phonetic, and 論 is a Chinese loanword so that might help, but by itself that's an even less common character, so you probably won't know it if you don't know 論. 侖 occurs as a phonetic in 倫 and 輪, but that's no help, because there it's "rin" instead of "ron". Did anyone guess "rin"?

"Sus" actually is "sub"! As in "look up to". From below, I guess, like the South Wind. Just as an overseer or supervisor sees from above rather than seeing something above him.

[thaumasiotes]

The word for lemon, 檸檬, that I mentioned above is a recent ateji. The characters come from some southern Chinese language borrowing the word "lemon" from English. The Mandarin pronunciation follows the characters, resulting in ningmeng, something just different enough that English speakers are unlikely to notice the connection.

The Japanese word for the fruit is also borrowed from English, resulting in the much closer レモン remon. But the spelling is borrowed from Chinese (when kanji are used at all), despite the pronunciation coming from elsewhere.

> There are kanji less obscure than 論. There are some that are so common that even I know them. It is in the 6th grade jōyō kanji, but I doubt it's in the top 1000 most common characters in Japanese, maybe not even the top 2000. You could probably guess that 侖 is the phonetic

It's not an especially common character in Chinese either, though it is part of some common words like 讨论. (论 is the simplified form of 論.) But frequency of occurrence isn't the only way you can fail to be obscure. I call this one of the least obscure characters in existence because it is the name of the most prominent Confucian work, the one conventionally titled Analects in English, 論語 "discussions".

Consider that "spangled" is not a word in common use in the USA, but Americans can be expected to recognize it anyway.

> Did anyone guess "rin"?

I don't remember.

> Presumably when prewett said "When the Japanese imported it, they used the characters much more phonetically," they meant onyomi. (Perhaps they didn't know about kunyomi.) When they say, "but the characters got assigned to the Japanese pronunciation of the word," that is mostly wrong as a description of how Japanese writing uses kanji today

I read prewett very differently than you do. To me, when he says "the characters got assigned to the Japanese pronunciation of the word", he is referring to kunyomi. I don't think that's wrong as a description of how Japanese writing uses kanji today. But I do think it's wrong to describe that use as "much more phonetic" than the Chinese writing system. I see it as much less phonetic.