Why is 左必之佐 read as サビシサ in 万葉集 entry 3734?

Question

Entry 3734 of the 万葉集 reads as follows¹:

等【ト】保【ホ】伎【キ】山【やま】 世【セ】伎【キ】毛【モ】故【コ】要【エ】伎【キ】奴【ヌ】 伊【イ】麻【マ】左【サ】良【ラ】尓【ニ】 安【ア】布【フ】倍【ベ】伎【キ】与【ヨ】之【シ】能【ノ】 奈【ナ】伎【キ】我【ガ】佐【サ】夫【ブ】之【シ】佐 【サ】

Transcribed in more modern Japanese style, this reads:

遠【とほ】き山【やま】関【せき】も越【こ】え来【き】ぬ今【いま】さらに逢【あ】ふべきよしのなきが寂【さぶ】しさ

It is noted that in certain 万葉集 transcriptions, the last four characters are transcribed as 左【サ】必【ビ】之【シ】佐 【サ】 （寂【さび】しさ）. However, what interests me about this alternate transcription is that the 万葉仮名 used for ビ is devoiced; generally it is read as 必【ヒ】. I do understand that, at least for the subsequent thousand years or so, the marking of voicing was inconsistent at best, but I was under the impression that in 上代特殊仮名遣【じょうだいとくしゅかなづかい】, a distinction between voiced and voiceless consonants was in use.

So why is this read as ビ? Am I wrong that a 清濁 distinction is in use generally within the 万葉集?

Note: incidentally, every other usage of 必【ヒ】 in the 万葉集 is devoiced. Sadly, the reference above is the only usage of サビシ in the 万葉集.

---

References

¹ http://etext.lib.virginia.edu/japanese/manyoshu/Man15Yo.html

---

Note: I originally mistakenly claimed that the 万葉仮名「夫」 was normally read as 夫【フ】. This is not true and I have updated the question to reflect this.

Answer 1

[Rewritten from comments above.]

Regarding the annotation 一云左必之佐, page 353 in 萬葉集三 (ISBN 4-00-240003-4) in the 新日本古典文学大系 series says:

西本願寺本などには「一云左必之佐」の注記があるが、天治本・類聚古集・広瀬本など非仙覚本系の諸本にはそれがない。また、「左必之佐」の「必」は甲類の仮名であるが、「‌​寂しさ」の意ならば「び」の仮名は乙類でなければならない。「一云左必之佐」の本文は、後世の竄入と思われる（有坂秀世『国語音韻史の研究』）。

As indicated, an analysis of the various manuscripts shows that this annotation does not appear in most of the manuscripts. More likely it was added at a later period by someone. Who wrote it and when is unknown, but speculation is possible. Several things can be noted:

• If sabisi derives from sabusi, the /bi/ must be 乙類.
• 必 is 甲類 rather than the expected 乙類.
• 必 is non-voiced /pi/ rather than the expected /bi/.
• 甲類 and 乙類 distinctions soon vanished after Old Japanese.
• While 万葉仮名 is able to distinguish between voiced consonants, that distinction is not always made. Further, early hiragana and katakana lacked a means to distinguish voice, and even after a method was developed, it was not regularly used like in Modern Japanese.

Considering these clues together with the hypothesis that this annotation was added at a later period, the following may be assumed:

• A later scribe would be unable to distinguish between 甲類 and 乙類 sounds since this distinction is now no longer made.
• The lack of voice distinctions may be normal a person of a later period.
• Perhaps this annotation represents a translation from the now old sabusi to the newer sabisi.

While much of this is supposition, in all likelihood this annotation was done by someone at in a later period.

Am I wrong that a 清濁 distinction is in use generally within the 万葉集?

In general, I would have to agree. However, there are exceptions even in 万葉集. Quote from 時代別国語大辞典：上代編, page 29: 万葉集では、清音仮名をもって濁音を写したと見なされる場合がややふえる（略）. These are not absolutes, so a bit of flexibility is needed at times. Note though that this point is rather moot now with the supposition that this above annotation was added by a later scribe so it does not represent the original 万葉集.