Theiling Online    Sitemap    Conlang Mailing List HQ   

Re: Japanese voicing, and learning Mandarin and Portuguese (was Re: Keyboards)

From:Nik Taylor <yonjuuni@...>
Date:Thursday, January 12, 2006, 2:42
Roger Mills wrote:
> This seems to require the speaker to examine the entire 2d lexeme before > deciding whether to voice or not-- probably an undue mental burden > especially if the governing vd.C is several syllables distant. No wonder > that kids learning to speak would be confused (and the wikipedia article > mentions that native speakers do indeed suffer confusion.) So nowadays > compds. with voicing have simply become lexicalized.
There is still a rule against voicing words that contain a voiced obstruent, e.g., tamago can never become *damago in compounds
> Was there ever, or is there now, any constraint in Japanese on lexemes with > consecutive syllables with vd. C ? e.g. /baga/, /guda/ etc? If so, that > might help explain it (though not cases of _V-vl-V-vd...).
Historically, there were two relevant rules: 1. No more than one voiced obstruent in a morpheme 2. No word may begin with a voiced obstruent There is a theory that rendaku originated by the addition of that second rule. E.g., _kata_ may have been _gata_ in ancient Japanese, but when the 2nd rule developed, it became _kata_. However, in compounds, it would've preserved the original /g/. Thus, if true, rendaku originated as a DEvoicing phenomenon. Later borrowings have introduced words with initial voiced obstruents, such as the _daku_ in _rendaku_. Furthermore, some native words have acquired voiced obstruents in Modern Japanese. For example, _dare_ ("who?") was _tare_ in Middle Japanese. (Incidentally, I'm not sure if this particular example is correct, or if -gata is an example of a later reanalysis, but the principle holds) Apparently in Old Japanese, words either did or did not become voiced in compounds, which is the justification for that theory. Later changes complicated the situation by the development of voicing in words with no etymological justification, or, vice versa, loss of voicing where it originally existed.
> And if perchance voicing takes place in a mixed bag of environments, that > might mean the rule must be written using braces { (curly brackets)-- and as > James McCawley always said (I paraphrase), "Curly brackets are evil" and > when you see them in a rule, you can safely bet there's something wrong. > :-)))
It seems that various borrowings, reanalyses, etc., have complicated the situation to a point where it's effectively a matter of memorizing when you voice and when you don't.