Rendaku (連濁 ‘sequential voicing’) refers in Japonic languages to a number of morphophonological alterations where the voiceless initial segment of non-initial element in a compound, under certain conditions, becomes its voiced counterpart.
A number of forms originated from historical contractions. For instance, there is a clear semantic difference between 山川 yamakawa [jamakawa] ‘mountains and rivers’ (a dvandva compound) and 山川 yamagawa [jamaɡawa] ‘mountain river’. Historically, yamagawa would have been a contraction of earlier *[jama=no kawa] |mountain=GEN river|. In earlier varieties of Japanese, voiced stops were in fact pre-nasalized voiced stops (cf. Miyake 2003: 196), so when the genitive case marker contracted, it would have contracted to a form like [jamaŋɡawa].
While these sorts of historical contractions produced rendaku alternations, a number of rules restrict when rendaku may occur (for a more full discussion, see Vance 1987: 133-148), and, additionally, a number of forms have been created through analogy.
For instance, 時々 tokidoki [tokidoki] ‘sometimes’ is historically a reduplication of the word 時 toki [toki] ‘time’. The initial segment of the second element is [d], the voiced counterpart to the original segment [t]. It is likely that no contraction took place in this form or any of the other reduplicated forms, but, based on an analogical extension of the rendaku pattern. these forms acquired a rendaku-like voicing alternation.
While there appear to be some examples of rendaku in Japonic languages other than Japanese, none of these are well studied at present. One example I noticed recently in Okinawan is the verbal auxiliary -busha- ‘to want to do (something)’, from the adjective husha- ‘to want, to wish for’. For instance:
(1) shumutsi husha-n
‘[I] want a book.’
(2) shumutsi kachi-busha-n
‘[I] want to write a book.’
Compare these with the standard Japanese cognate 欲しい hoshii ‘to want, to wish for’ and the auxiliary ほしい hoshii:
(3) hon=ga hoshi-i
‘[I] want a book.’
(4) hon=o ka-ite hoshi-i
book=ACC write-SUB want-IPFV
‘[I] want [someone else] to write a book.’
In addition to the rendaku alternation in Okinawan, the auxiliary co-occurs with the infinitive form of the verb. In Japanese, there is no rendaku alternation, and this auxiliary co-occurs with with the subordinating converb -(i)te, rather than the infinitive. And there is a semantic difference between the Okinawan and Japanese auxiliary: the Okinawan auxiliary busha- means that speaker wants to do something, while the Japanese auxiliary hoshi- means that the speaker wants someone else to do something. The Japanese equivalent to (2) would use the auxiliary たい tai:
(5) hon=o kak-i-ta-i
‘[I] want to write a book.’
Okinawan busha- is apparently a case of analogical rendaku, as there should be no intervening elements, even historically, between the infinitive and the auxiliary. So what about contraction-based rendaku? This appears to be more sparse in Okinawan than in Japanese, but still does appear to occur. For instance, the Okinawan word for ‘hair’ is kii [kiː], and we find several forms where an apparent contraction-motivated rendaku has occured. For instance, one word for ‘hair (on the head)’ is karazigii [kaɾadʑiɡiː], likely from *karazi=nu kii |head.hair=GEN hair| (this uncontracted form, [kaɾadʑinukiː], is still extant in Okinawan). Similarly, ‘nose hair’ is hanagii [hanaɡiː], likely from *hana=nu kii.
However, the exact situation in Okinawan is complicated by the large number of loans from Japanese, some which were likely borrowed as rendaku. This can make it difficult to untangle the exact extent of rendaku in Okinawan. For instance, Okinawan has borrowed the Japanese word tamago [tamaɡo] ‘egg’ as tamagu [tamaɡu] ‘egg’, while the native Okinawan word is kuuga [kuuɡa] ‘egg’. This Japanese form is from the compound *tama=no ko |ball=GEN small.thing|, with a rendaku alternation of ko ‘child, egg, small thing’. Thus, while Okinawan tamagu does come from a rendaku alternation, it is likely not a result of rendaku in Okinawan.
I think one likely productive area of research would be on the various sound symbolic words in Okinawan. These appear to be quite resistant to borrowing, and most are distinct from their Japanese equivalents. For instance, Japanese kankan ‘[shine] intensely’ is equivalent to Okinawan kwankwan ‘[shine] intensely’. On quick inspection, these sound symbolic words in Okinawan seem to conform to the same rendaku rules that Japanese does.
Further attention, as always, is warranted.