I’ve noticed that kun’yomi words that contain a rendaku-able first kana don’t get included in compounds, or don’t get rendaku in compounds, if the second or above kana has a rendaku. I.e. all of these words ひざ・食べる・外す・恥ずかしい could become びざ・だべる・ばずす・ばずかしい in theory in some kind of compound. But based on my observations, and also checking a few words on [jisho.org](https://jisho.org) (for example looking up ＊だべ and ＊だべる and getting no results), it pretty much never happens.
Is this a rule? If so I think it would make a lot of sense, because the rendaku represents a kind of break in the word where you could pick apart the roots, and this kind of breaks it… Anyway, I could see how it would come from the way rendaku originally worked with how I understand it. I was curious if anyone knew the answer or had a counterexample, though. Thanks for reading through this
1 comment
Yes, this is a rule and it’s called “Lyman’s Law”. Timothy Vance in his chapter on Rendaku in the “Handbook of Japanese Phonetics and Phonology” states it as “A non-initial voiced obstruent in a morph inhibits rendaku”.
A famous counter example for Lyman’s Law is 縄梯子 (なわばしご). Given the word consists of なわ and はしご, with the latter one containing a non-initial voiced obstruent, it should be immune to rendaku but isn’t.
Note that the definition mentions the second word having a voice obstruent rather than having rendaku. The べ in 食べる for example, is just a voiced consonant and not considered rendaku as 食べる is not the compound た＋べる.
See also https://en.wikipedia.org/wiki/Rendaku#Lyman's_Law