About ancient pronunciation on dictionaries

Question

In large dictionaries like 広辞苑 or 大辞林, there are some entries tagged with "ancient pronunciations" which are almost identical to the modern ones but with a voiced consonant voiceless (or vice versa), e.g.

さわ・ぐ 【騒ぐ】〔上代は「さわく」と清音。〕pronounced voicelessly as "sawaku" in the early era

しろ-がね 【銀】〔古くは「しろかね」とも。〕also "shirokane" in ancient times

As I know, in most ancient texts, voiced syllables are written without any voicing mark(濁点), so they are identical to the voiceless ones orthographically. How do dictionary editors know that the syllables are with voiceless consonants?

Answer 1

There are at least two ways that the difference between "voiceless" and "voiced"¹ pronunciation can be ascertained.

First, through the 日葡辞書, the Japanese-Portuguese dictionary of 1603, which uses Latin script to render Japanese. You can hardly argue for しろがね, when it literally has the entry "Xirocane, prata" but no *xirogane.

Furthermore, the Old Japanese man'yōgana, while definitely able to conflate the voiceless-voiced pairs (and occasionally the voiced-nasal pairs), is usually consistent in this regard. For example, in the usual orthohraphy of the Man'yōshū , く is spelt with characters such as 久玖口群苦丘九鳩君(on)來國小(kun), while ぐ by 具遇求隅群(on)來(kun). As we see, while there are two cases of conflation, most are distinguished, and when we see, say, in 14.3349, written fully phonetically, 布奈妣等佐和久 <punabîtö sawaku>, this 久 makes us sure we are looking at a く and not a ぐ.

¹ Well, actually "fortis" and "lenis", as the distinction between the two did not originally involve voicedness, but prenazalization, and in Tōhoku this remains still.

Answer 2

Lexicographers today can differentiate voiced/voiceless from Manyogana patterns in the ancient text. However, I believe that is a relatively new linguistic finding made by researchers like Motoori Norinaga. I don't know what lexicographers from earlier periods did.

In an edition of 倭玉篇 (1613), Section 237 for 金, 銅 is アカゞネ ("akagane") while 銀 is シロカネ ("shirokane"). In this case, the author could have written シロガネ but didn't. There was an earlier, lost edition from 1489 (長享本), and we don't know how the words were written there, though.

I believe there are words proven to be written in Manyogana, and those written with dakuten (as in 倭玉篇), among "old" texts. Perhaps the lexical items found in neither of the two categories might not be as many as a layperson might think.